fix: CUDA device PCI bus ID de-dupe OOMing (ignoring other 3 gpus entirely) (#22533)

* fix: CUDA device PCI bus ID detection for multi-GPU de-dupe

* HIP, MUSA macros

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>

.devops/intel.Dockerfile

@@ -1,4 +1,4 @@
-ARG ONEAPI_VERSION=2025.3.2-0-devel-ubuntu24.04
+ARG ONEAPI_VERSION=2025.3.3-0-devel-ubuntu24.04
 
 ## Build Image
 

.devops/nix/package.nix

@@ -18,6 +18,7 @@
   vulkan-loader,
   openssl,
   shaderc,
+  spirv-headers,
   useBlas ?
     builtins.all (x: !x) [
       useCuda
@@ -145,6 +146,7 @@ effectiveStdenv.mkDerivation (finalAttrs: {
       ninja
       pkg-config
       git
+      spirv-headers
     ]
     ++ optionals useCuda [
       cudaPackages.cuda_nvcc

.devops/openvino.Dockerfile

@@ -2,7 +2,19 @@ ARG OPENVINO_VERSION_MAJOR=2026.0
 ARG OPENVINO_VERSION_FULL=2026.0.0.20965.c6d6a13a886
 ARG UBUNTU_VERSION=24.04
 
-# Optional proxy build arguments - empty by default
+# Intel GPU driver versions. https://github.com/intel/compute-runtime/releases
+ARG IGC_VERSION=v2.30.1
+ARG IGC_VERSION_FULL=2_2.30.1+20950
+ARG COMPUTE_RUNTIME_VERSION=26.09.37435.1
+ARG COMPUTE_RUNTIME_VERSION_FULL=26.09.37435.1-0
+ARG IGDGMM_VERSION=22.9.0
+
+# Intel NPU driver versions. https://github.com/intel/linux-npu-driver/releases
+ARG NPU_DRIVER_VERSION=v1.32.0
+ARG NPU_DRIVER_FULL=v1.32.0.20260402-23905121947
+ARG LIBZE1_VERSION=1.27.0-1~24.04~ppa2
+
+# Optional proxy build arguments
 ARG http_proxy=
 ARG https_proxy=
 
@@ -78,13 +90,47 @@ ARG http_proxy
 ARG https_proxy
 
 RUN apt-get update \
-    && apt-get install -y libgomp1 libtbb12 curl \
+    && apt-get install -y libgomp1 libtbb12 curl wget ocl-icd-libopencl1 \
     && apt autoremove -y \
     && apt clean -y \
     && rm -rf /tmp/* /var/tmp/* \
     && find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
     && find /var/cache -type f -delete
 
+# Install GPU drivers
+ARG IGC_VERSION
+ARG IGC_VERSION_FULL
+ARG COMPUTE_RUNTIME_VERSION
+ARG COMPUTE_RUNTIME_VERSION_FULL
+ARG IGDGMM_VERSION
+RUN mkdir /tmp/neo/ && cd /tmp/neo/ \
+    && wget https://github.com/intel/intel-graphics-compiler/releases/download/${IGC_VERSION}/intel-igc-core-${IGC_VERSION_FULL}_amd64.deb \
+    && wget https://github.com/intel/intel-graphics-compiler/releases/download/${IGC_VERSION}/intel-igc-opencl-${IGC_VERSION_FULL}_amd64.deb \
+    && wget https://github.com/intel/compute-runtime/releases/download/${COMPUTE_RUNTIME_VERSION}/intel-ocloc-dbgsym_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.ddeb \
+    && wget https://github.com/intel/compute-runtime/releases/download/${COMPUTE_RUNTIME_VERSION}/intel-ocloc_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.deb \
+    && wget https://github.com/intel/compute-runtime/releases/download/${COMPUTE_RUNTIME_VERSION}/intel-opencl-icd-dbgsym_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.ddeb \
+    && wget https://github.com/intel/compute-runtime/releases/download/${COMPUTE_RUNTIME_VERSION}/intel-opencl-icd_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.deb \
+    && wget https://github.com/intel/compute-runtime/releases/download/${COMPUTE_RUNTIME_VERSION}/libigdgmm12_${IGDGMM_VERSION}_amd64.deb \
+    && wget https://github.com/intel/compute-runtime/releases/download/${COMPUTE_RUNTIME_VERSION}/libze-intel-gpu1-dbgsym_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.ddeb \
+    && wget https://github.com/intel/compute-runtime/releases/download/${COMPUTE_RUNTIME_VERSION}/libze-intel-gpu1_${COMPUTE_RUNTIME_VERSION_FULL}_amd64.deb \
+    && dpkg --install *.deb \
+    && rm -rf /tmp/neo/
+
+# Install NPU drivers
+ARG NPU_DRIVER_VERSION
+ARG NPU_DRIVER_FULL
+ARG LIBZE1_VERSION
+RUN mkdir /tmp/npu/ && cd /tmp/npu/ \
+    && wget https://github.com/intel/linux-npu-driver/releases/download/${NPU_DRIVER_VERSION}/linux-npu-driver-${NPU_DRIVER_FULL}-ubuntu2404.tar.gz \
+    && tar -xf linux-npu-driver-${NPU_DRIVER_FULL}-ubuntu2404.tar.gz \
+    && dpkg --install *.deb \
+    && rm -rf /tmp/npu/
+
+RUN cd /tmp \
+    && wget https://snapshot.ppa.launchpadcontent.net/kobuk-team/intel-graphics/ubuntu/20260324T100000Z/pool/main/l/level-zero-loader/libze1_${LIBZE1_VERSION}_amd64.deb \
+    && dpkg --install libze1_${LIBZE1_VERSION}_amd64.deb \
+    && rm libze1_${LIBZE1_VERSION}_amd64.deb
+
 COPY --from=build /app/lib/ /app/
 
 ### Full (all binaries)

.github/ISSUE_TEMPLATE/010-bug-compilation.yml

@@ -12,6 +12,8 @@ body:
         after recreating the CMake build directory and with `-DGGML_CCACHE=OFF`.
         If the compilation succeeds with ccache disabled you should be able to permanently fix the issue
         by clearing `~/.cache/ccache` (on Linux).
+
+        Please fill out this template yourself, copypasting language model outputs is [strictly prohibited](https://github.com/ggml-org/llama.cpp/blob/master/CONTRIBUTING.md#ai-usage-policy).
   - type: textarea
     id: commit
     attributes:

.github/ISSUE_TEMPLATE/011-bug-results.yml

@@ -1,5 +1,5 @@
 name: Bug (model use)
-description: Something goes wrong when using a model (in general, not specific to a single llama.cpp module).
+description: Something goes wrong when running a model (crashes, garbled outputs, etc.).
 title: "Eval bug: "
 labels: ["bug-unconfirmed", "model evaluation"]
 body:
@@ -12,6 +12,8 @@ body:
         If you encountered the issue while using an external UI (e.g. ollama),
         please reproduce your issue using one of the examples/binaries in this repository.
         The `llama-completion` binary can be used for simple and reproducible model inference.
+
+        Please fill out this template yourself, copypasting language model outputs is [strictly prohibited](https://github.com/ggml-org/llama.cpp/blob/master/CONTRIBUTING.md#ai-usage-policy).
   - type: textarea
     id: version
     attributes:

.github/ISSUE_TEMPLATE/019-bug-misc.yml

@@ -10,6 +10,8 @@ body:
         This issue template is intended for miscellaneous bugs that don't fit into any other category.
         If you encountered the issue while using an external UI (e.g. ollama),
         please reproduce your issue using one of the examples/binaries in this repository.
+
+        Please fill out this template yourself, copypasting language model outputs is [strictly prohibited](https://github.com/ggml-org/llama.cpp/blob/master/CONTRIBUTING.md#ai-usage-policy).
   - type: textarea
     id: version
     attributes:

.github/ISSUE_TEMPLATE/020-enhancement.yml

@@ -8,6 +8,8 @@ body:
       value: |
         [Please post your idea first in Discussion if there is not yet a consensus for this enhancement request. This will help to keep this issue tracker focused on enhancements that the community has agreed needs to be implemented.](https://github.com/ggml-org/llama.cpp/discussions/categories/ideas)
 
+        Please fill out this template yourself, copypasting language model outputs is [strictly prohibited](https://github.com/ggml-org/llama.cpp/blob/master/CONTRIBUTING.md#ai-usage-policy).
+
   - type: checkboxes
     id: prerequisites
     attributes:

.github/ISSUE_TEMPLATE/030-research.yml

@@ -8,6 +8,8 @@ body:
       value: |
         Don't forget to check for any [duplicate research issue tickets](https://github.com/ggml-org/llama.cpp/issues?q=is%3Aopen+is%3Aissue+label%3A%22research+%F0%9F%94%AC%22)
 
+        Please fill out this template yourself, copypasting language model outputs is [strictly prohibited](https://github.com/ggml-org/llama.cpp/blob/master/CONTRIBUTING.md#ai-usage-policy).
+
   - type: checkboxes
     id: research-stage
     attributes:

.github/ISSUE_TEMPLATE/040-refactor.yml

@@ -9,6 +9,8 @@ body:
         Don't forget to [check for existing refactor issue tickets](https://github.com/ggml-org/llama.cpp/issues?q=is%3Aopen+is%3Aissue+label%3Arefactoring) in case it's already covered.
         Also you may want to check [Pull request refactor label as well](https://github.com/ggml-org/llama.cpp/pulls?q=is%3Aopen+is%3Apr+label%3Arefactoring) for duplicates too.
 
+        Please fill out this template yourself, copypasting language model outputs is [strictly prohibited](https://github.com/ggml-org/llama.cpp/blob/master/CONTRIBUTING.md#ai-usage-policy).
+
   - type: textarea
     id: background-description
     attributes:

.github/pull_request_template.md

@@ -6,7 +6,7 @@
 
 <!-- You can provide more details and link related discussions here. Delete this section if not applicable -->
 
-# Requirements
+## Requirements
 
 <!-- IMPORTANT: Please do NOT delete this section, otherwise your PR may be rejected -->
 

.github/workflows/build-and-test-snapdragon.yml

@@ -0,0 +1,116 @@
+name: CI (snapdragon)
+
+on:
+  workflow_dispatch:
+  push:
+    branches:
+      - master
+    paths:
+      - '.github/workflows/build-and-test-snapdragon.yml'
+      - 'ggml/include/ggml-hexagon.h'
+      - 'ggml/src/ggml-hexagon/**'
+      - 'docs/backend/snapdragon/**'
+      - 'scripts/snapdragon/**'
+      - 'CMakePresets.json'
+
+  pull_request:
+    types: [opened, synchronize, reopened]
+    paths:
+      - '.github/workflows/build-and-test-snapdragon.yml'
+      - 'ggml/include/ggml-hexagon.h'
+      - 'ggml/src/ggml-hexagon/**'
+      - 'docs/backend/snapdragon/**'
+      - 'scripts/snapdragon/**'
+      - 'CMakePresets.json'
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
+  cancel-in-progress: true
+
+jobs:
+  android-ndk-snapdragon:
+    runs-on: ubuntu-latest
+    container:
+      image: 'ghcr.io/snapdragon-toolchain/arm64-android:v0.3'
+    defaults:
+      run:
+        shell: bash
+
+    steps:
+      - name: Clone
+        uses: actions/checkout@v6
+        with:
+          fetch-depth: 0
+          lfs: false
+
+      - name: Build Llama.CPP for Snapdragon Android
+        id: build_llama_cpp_snapdragon_android
+        run: |
+          cp docs/backend/snapdragon/CMakeUserPresets.json .
+          cmake --preset arm64-android-snapdragon-release -B build
+          cmake --build build
+          cmake --install build --prefix pkg-snapdragon/llama.cpp
+
+      - name: Upload Llama.CPP Snapdragon Android Build Artifact
+        if: ${{ always() && steps.build_llama_cpp_snapdragon_android.outcome == 'success' }}
+        uses: actions/upload-artifact@v6
+        with:
+          name: llama-cpp-android-arm64-snapdragon
+          path: pkg-snapdragon/llama.cpp
+
+  test-snapdragon-qdc:
+    name: Test on QDC Android Device (${{ matrix.device }})
+    needs: [android-ndk-snapdragon]
+    runs-on: ubuntu-slim
+    strategy:
+      fail-fast: false
+      matrix:
+        device: [SM8750, SM8650, SM8850]
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+
+      - name: Download build artifact
+        uses: actions/download-artifact@v7
+        with:
+          name: llama-cpp-android-arm64-snapdragon
+          path: pkg-snapdragon/llama.cpp
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.x'
+          cache: pip
+
+      - name: Install system dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y curl unzip
+
+      - name: Install QDC SDK wheel
+        run: |
+          curl -fSL -o qdc_sdk.zip https://softwarecenter.qualcomm.com/api/download/software/tools/Qualcomm_Device_Cloud_SDK/All/0.2.3/qualcomm_device_cloud_sdk-0.2.3.zip
+          unzip qdc_sdk.zip -d qdc_sdk
+          pip install qdc_sdk/qualcomm_device_cloud_sdk-0.2.3-py3-none-any.whl
+
+      - name: Check QDC API key
+        id: check_secret
+        env:
+          QDC_API_KEY: ${{ secrets.QDC_API_KEY }}
+        run: echo "has-qdc-key=${{ env.QDC_API_KEY != '' }}" >> "$GITHUB_OUTPUT"
+
+      - name: Run QDC tests (${{ matrix.device }})
+        if: steps.check_secret.outputs.has-qdc-key == 'true'
+        run: |
+          python scripts/snapdragon/qdc/run_qdc_jobs.py \
+              --test       all \
+              --pkg-dir    pkg-snapdragon/llama.cpp \
+              --model-url  "https://huggingface.co/bartowski/Llama-3.2-1B-Instruct-GGUF/resolve/main/Llama-3.2-1B-Instruct-Q4_0.gguf" \
+              --device     ${{ matrix.device }}
+        env:
+          QDC_API_KEY: ${{ secrets.QDC_API_KEY }}
+
+      - name: Cleanup
+        if: always()
+        run: rm -rf pkg-snapdragon qdc_sdk qdc_sdk.zip

.github/workflows/build-android.yml

@@ -1,26 +1,24 @@
 name: CI (android)
 
 on:
-  workflow_dispatch: # allows manual triggering
+  workflow_dispatch:
   push:
     branches:
       - master
-    paths: [
-      '.github/workflows/build-android.yml',
-      '**/CMakeLists.txt',
-      '**/.cmake',
-      '**/*.h',
-      '**/*.hpp',
-      '**/*.c',
-      '**/*.cpp'
-    ]
+    paths:
+      - '.github/workflows/build-android.yml'
+      - '**/CMakeLists.txt'
+      - '**/.cmake'
+      - '**/*.h'
+      - '**/*.hpp'
+      - '**/*.c'
+      - '**/*.cpp'
 
   pull_request:
     types: [opened, synchronize, reopened]
-    paths: [
-      '.github/workflows/build-android.yml',
-      'examples/llama.android/**'
-    ]
+    paths:
+      - '.github/workflows/build-android.yml'
+      - 'examples/llama.android/**'
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
@@ -51,7 +49,7 @@ jobs:
           distribution: zulu
 
       - name: Setup Android SDK
-        uses: android-actions/setup-android@9fc6c4e9069bf8d3d10b2204b1fb8f6ef7065407 # v3
+        uses: android-actions/setup-android@40fd30fb8d7440372e1316f5d1809ec01dcd3699 # v4.0.1
         with:
           log-accepted-android-sdk-licenses: false
 
@@ -67,35 +65,24 @@ jobs:
     defaults:
       run:
         shell: bash
-    strategy:
-      matrix:
-        include:
-          - build: 'arm64-cpu'
-            defines: '-D ANDROID_ABI=arm64-v8a -D ANDROID_PLATFORM=android-31 -D CMAKE_TOOLCHAIN_FILE=${ANDROID_NDK_ROOT}/build/cmake/android.toolchain.cmake -D GGML_NATIVE=OFF -DGGML_CPU_ARM_ARCH=armv8.5-a+fp16+i8mm -G Ninja -D LLAMA_OPENSSL=OFF -D GGML_OPENMP=OFF'
-          - build: 'arm64-snapdragon'
-            defines: '--preset arm64-android-snapdragon-release'
 
     steps:
       - name: Clone
-        id: checkout
         uses: actions/checkout@v6
         with:
           fetch-depth: 0
           lfs: false
 
-      - name: Build Llama.CPP for Hexagon Android
-        id: build_llama_cpp_hexagon_android
+      - name: Build
+        id: ndk_build
         run: |
-          if [[ "${{ matrix.build }}" == "arm64-snapdragon" ]]; then
-            cp docs/backend/snapdragon/CMakeUserPresets.json .
-          fi
-          cmake ${{ matrix.defines }} -B build
+          cmake -D ANDROID_ABI=arm64-v8a -D ANDROID_PLATFORM=android-31 -D CMAKE_TOOLCHAIN_FILE=${ANDROID_NDK_ROOT}/build/cmake/android.toolchain.cmake -D GGML_NATIVE=OFF -DGGML_CPU_ARM_ARCH=armv8.5-a+fp16+i8mm -G Ninja -D LLAMA_OPENSSL=OFF -D GGML_OPENMP=OFF -B build
           cmake --build build
           cmake --install build --prefix pkg-adb/llama.cpp
 
-      - name: Upload Llama.CPP Hexagon Android Build Artifact
-        if: ${{ always() && steps.build_llama_cpp_hexagon_android.outcome == 'success' }}
+      - name: Upload Android Build Artifact
+        if: ${{ always() && steps.ndk_build.outcome == 'success' }}
         uses: actions/upload-artifact@v6
         with:
-          name: llama-cpp-android-${{ matrix.build }}
+          name: llama-cpp-android-arm64-cpu
           path: pkg-adb/llama.cpp

.github/workflows/build-cross.yml

@@ -246,6 +246,7 @@ jobs:
           apt-get install -y --no-install-recommends \
                   build-essential \
                   glslc \
+                  spirv-headers \
                   gcc-14-loongarch64-linux-gnu \
                   g++-14-loongarch64-linux-gnu \
                   libvulkan-dev:loong64

.github/workflows/build-openvino.yml

@@ -0,0 +1,120 @@
+name: CI (openvino)
+
+on:
+  workflow_dispatch: # allows manual triggering
+  push:
+    branches:
+      - master
+    paths: [
+      '.github/workflows/build-openvino.yml',
+      '**/CMakeLists.txt',
+      '**/.cmake',
+      '**/*.h',
+      '**/*.hpp',
+      '**/*.c',
+      '**/*.cpp',
+    ]
+
+  pull_request:
+    types: [opened, synchronize, reopened]
+    paths: [
+      '.github/workflows/build-openvino.yml',
+      'ggml/src/ggml-openvino/**'
+    ]
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
+  cancel-in-progress: true
+
+env:
+  GGML_NLOOP: 3
+  GGML_N_THREADS: 1
+  LLAMA_LOG_COLORS: 1
+  LLAMA_LOG_PREFIX: 1
+  LLAMA_LOG_TIMESTAMPS: 1
+
+jobs:
+  ubuntu-24-openvino:
+    name: ubuntu-24-openvino-${{ matrix.openvino_device }}
+
+    concurrency:
+      group: openvino-${{ matrix.variant }}-${{ github.head_ref || github.ref }}
+      cancel-in-progress: false
+
+    strategy:
+      matrix:
+        include:
+          - variant: cpu
+            runner: '"ubuntu-24.04"'
+            openvino_device: "CPU"
+          - variant: gpu
+            runner: '["self-hosted","Linux","Intel","OpenVINO"]'
+            openvino_device: "GPU"
+
+    runs-on: ${{ fromJSON(matrix.runner) }}
+
+    env:
+      # Sync versions in build-openvino.yml, build-self-hosted.yml, release.yml, build-cache.yml, .devops/openvino.Dockerfile
+      OPENVINO_VERSION_MAJOR: "2026.0"
+      OPENVINO_VERSION_FULL: "2026.0.0.20965.c6d6a13a886"
+
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v6
+
+      - name: ccache
+        if: runner.environment == 'github-hosted'
+        uses: ggml-org/ccache-action@v1.2.21
+        with:
+          key: ubuntu-24-openvino-${{ matrix.variant }}-no-preset-v1
+          evict-old-files: 1d
+          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+
+      - name: Dependencies
+        id: depends
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y build-essential libssl-dev libtbb12 cmake ninja-build python3-pip
+          sudo apt-get install -y ocl-icd-opencl-dev opencl-headers opencl-clhpp-headers intel-opencl-icd
+
+      - name: Use OpenVINO Toolkit Cache
+        if: runner.environment == 'github-hosted'
+        uses: actions/cache@v5
+        id: cache-openvino
+        with:
+          path: ./openvino_toolkit
+          key: openvino-toolkit-v${{ env.OPENVINO_VERSION_FULL }}-${{ runner.os }}
+
+      - name: Setup OpenVINO Toolkit
+        if: steps.cache-openvino.outputs.cache-hit != 'true'
+        uses: ./.github/actions/linux-setup-openvino
+        with:
+          path: ./openvino_toolkit
+          version_major: ${{ env.OPENVINO_VERSION_MAJOR }}
+          version_full: ${{ env.OPENVINO_VERSION_FULL }}
+
+      - name: Install OpenVINO dependencies
+        run: |
+          cd ./openvino_toolkit
+          chmod +x ./install_dependencies/install_openvino_dependencies.sh
+          echo "Y" | sudo -E ./install_dependencies/install_openvino_dependencies.sh
+
+      - name: Build
+        id: cmake_build
+        run: |
+          source ./openvino_toolkit/setupvars.sh
+          cmake -B build/ReleaseOV -G Ninja \
+            -DCMAKE_BUILD_TYPE=Release \
+            -DGGML_OPENVINO=ON
+          time cmake --build build/ReleaseOV --config Release -j $(nproc)
+
+      - name: Test
+        id: cmake_test
+        # TODO: fix and re-enable the `test-llama-archs` test below
+        run: |
+          cd ${{ github.workspace }}
+          if [ "${{ matrix.openvino_device }}" = "GPU" ]; then
+            export GGML_OPENVINO_DEVICE=GPU
+          fi
+          ctest --test-dir build/ReleaseOV -L main -E "test-llama-archs" --verbose --timeout 2000

.github/workflows/build-riscv.yml

@@ -47,22 +47,10 @@ jobs:
     steps:
       - name: Install dependencies
         run: |
-          sudo apt-get update
-
-          # Install necessary packages
-          sudo apt-get install -y libatomic1 libtsan2 gcc-14 g++-14 cmake build-essential wget git-lfs
-
           # Set gcc-14 and g++-14 as the default compilers
           sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-14 100
           sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-14 100
 
-          if ! which rustc; then
-            # Install Rust stable version
-            sudo apt-get install -y rustup
-            rustup install stable
-            rustup default stable
-          fi
-
           git lfs install
 
       - name: GCC version check
@@ -74,12 +62,12 @@ jobs:
         id: checkout
         uses: actions/checkout@v6
 
-      # FIXME: Enable when ggml-org/ccache-action works on riscv64
-      # - name: ccache
-      #   uses: ggml-org/ccache-action@v1.2.21
-      #   with:
-      #     key: ubuntu-riscv64-native-sanitizer-${{ matrix.sanytizer }}-${{ matrix.build_type }}
-      #     save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+      - name: ccache
+        uses: ggml-org/ccache-action@afde29e5b5422e5da23cb1f639e8baecadeadfc3 # https://github.com/ggml-org/ccache-action/pull/1
+        with:
+          key: ubuntu-riscv64-native-sanitizer-${{ matrix.sanitizer }}-${{ matrix.build_type }}
+          evict-old-files: 1d
+          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
 
       - name: Build
         id: cmake_build

.github/workflows/build-self-hosted.yml

@@ -97,6 +97,36 @@ jobs:
           vulkaninfo --summary
           GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
 
+  # TODO: investigate slight precision issues in some operations for test-backend-ops on the WebGPU backend.
+  #ggml-ci-nvidia-webgpu:
+  #  runs-on: [self-hosted, Linux, NVIDIA]
+
+  #  steps:
+  #    - name: Clone
+  #      id: checkout
+  #      uses: actions/checkout@v6
+
+  #    - name: Dawn Dependency
+  #      id: dawn-depends
+  #      run: |
+  #        DAWN_VERSION="v20260317.182325"
+  #        DAWN_OWNER="google"
+  #        DAWN_REPO="dawn"
+  #        DAWN_ASSET_NAME="Dawn-18eb229ef5f707c1464cc581252e7603c73a3ef0-ubuntu-latest-Release"
+  #        echo "Fetching release asset from https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
+  #        curl -L -o artifact.tar.gz \
+  #          "https://github.com/google/dawn/releases/download/${DAWN_VERSION}/${DAWN_ASSET_NAME}.tar.gz"
+  #        mkdir dawn
+  #        tar -xvf artifact.tar.gz -C dawn --strip-components=1
+
+  #    - name: Test
+  #      id: ggml-ci
+  #      run: |
+  #        GG_BUILD_WEBGPU=1 \
+  #        GG_BUILD_WEBGPU_DAWN_PREFIX="$GITHUB_WORKSPACE/dawn" \
+  #        GG_BUILD_WEBGPU_DAWN_DIR="$GITHUB_WORKSPACE/dawn/lib64/cmake/Dawn" \
+  #          bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
+
   # TODO: provision AMX-compatible machine
   #ggml-ci-cpu-amx:
   #  runs-on: [self-hosted, Linux, CPU, AMX]
@@ -235,6 +265,10 @@ jobs:
   ggml-ci-intel-openvino-gpu-low-perf:
     runs-on: [self-hosted, Linux, Intel, OpenVINO]
 
+    concurrency:
+      group: openvino-gpu-${{ github.head_ref || github.ref }}
+      cancel-in-progress: false
+
     env:
       # Sync versions in build.yml, build-self-hosted.yml, release.yml, build-cache.yml, .devops/openvino.Dockerfile
       OPENVINO_VERSION_MAJOR: "2026.0"

.github/workflows/build-sycl.yml

@@ -0,0 +1,142 @@
+name: CI (sycl)
+
+on:
+  workflow_dispatch: # allows manual triggering
+  push:
+    branches:
+      - master
+    paths: [
+      '.github/workflows/build-sycl.yml',
+      '**/CMakeLists.txt',
+      '**/.cmake',
+      '**/*.h',
+      '**/*.hpp',
+      '**/*.c',
+      '**/*.cpp'
+    ]
+
+  pull_request:
+    types: [opened, synchronize, reopened]
+    paths: [
+      '.github/workflows/build-sycl.yml',
+      'ggml/src/ggml-sycl/**'
+    ]
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
+  cancel-in-progress: true
+
+env:
+  GGML_NLOOP: 3
+  GGML_N_THREADS: 1
+  LLAMA_LOG_COLORS: 1
+  LLAMA_LOG_PREFIX: 1
+  LLAMA_LOG_TIMESTAMPS: 1
+
+jobs:
+
+  ubuntu-24-sycl:
+    strategy:
+      matrix:
+        build: [fp32, fp16]
+        include:
+          - build: fp32
+            fp16: OFF
+          - build: fp16
+            fp16: ON
+
+    runs-on: ubuntu-24.04
+
+    env:
+      ONEAPI_ROOT: /opt/intel/oneapi/
+      ONEAPI_INSTALLER_VERSION: "2025.3.3"
+
+    continue-on-error: true
+
+    steps:
+      - uses: actions/checkout@v6
+
+      - name: Use oneAPI Installation Cache
+        uses: actions/cache@v5
+        id: cache-sycl
+        with:
+          path: ${{ env.ONEAPI_ROOT }}
+          key: oneAPI-${{ env.ONEAPI_INSTALLER_VERSION }}-${{ runner.os }}
+
+      - name: Download & Install oneAPI
+        shell: bash
+        if: steps.cache-sycl.outputs.cache-hit != 'true'
+        run: |
+          cd /tmp
+          wget https://registrationcenter-download.intel.com/akdlm/IRC_NAS/56f7923a-adb8-43f3-8b02-2b60fcac8cab/intel-deep-learning-essentials-2025.3.3.16_offline.sh -O intel-deep-learning-essentials_offline.sh
+          sudo bash intel-deep-learning-essentials_offline.sh -s -a --silent --eula accept
+
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v6
+
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.21
+        with:
+          key: ubuntu-24-sycl-${{ matrix.build }}
+          evict-old-files: 1d
+          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+
+      - name: Build
+        id: cmake_build
+        run: |
+          source /opt/intel/oneapi/setvars.sh
+          cmake -B build \
+            -G "Ninja" \
+            -DCMAKE_BUILD_TYPE=Release \
+            -DGGML_SYCL=ON \
+            -DCMAKE_C_COMPILER=icx \
+            -DCMAKE_CXX_COMPILER=icpx \
+            -DLLAMA_OPENSSL=OFF \
+            -DGGML_NATIVE=OFF \
+            -DGGML_SYCL_F16=${{ matrix.fp16 }}
+          time cmake --build build --config Release -j $(nproc)
+
+  windows-latest-sycl:
+    runs-on: windows-2022
+
+    defaults:
+      run:
+        shell: bash
+
+    env:
+      WINDOWS_BASEKIT_URL: https://registrationcenter-download.intel.com/akdlm/IRC_NAS/b60765d1-2b85-4e85-86b6-cb0e9563a699/intel-deep-learning-essentials-2025.3.3.18_offline.exe
+      WINDOWS_DPCPP_MKL: intel.oneapi.win.cpp-dpcpp-common:intel.oneapi.win.mkl.devel:intel.oneapi.win.dnnl:intel.oneapi.win.tbb.devel
+      ONEAPI_ROOT: "C:/Program Files (x86)/Intel/oneAPI"
+      ONEAPI_INSTALLER_VERSION: "2025.3.3"
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v6
+
+      - name: Use oneAPI Installation Cache
+        uses: actions/cache@v5
+        id: cache-sycl
+        with:
+          path: ${{ env.ONEAPI_ROOT }}
+          key: oneAPI-${{ env.ONEAPI_INSTALLER_VERSION }}-${{ runner.os }}
+
+      - name: Download & Install oneAPI
+        shell: bash
+        if: steps.cache-sycl.outputs.cache-hit != 'true'
+        run: |
+          scripts/install-oneapi.bat $WINDOWS_BASEKIT_URL $WINDOWS_DPCPP_MKL
+
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.21
+        with:
+          key: windows-latest-sycl
+          variant: ccache
+          evict-old-files: 1d
+          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+
+      # TODO: add ssl support ; we will also need to modify win-build-sycl.bat to accept user-specified args
+
+      - name: Build
+        id: cmake_build
+        run:  examples/sycl/win-build-sycl.bat

.github/workflows/build.yml

@@ -267,6 +267,56 @@ jobs:
           wget https://huggingface.co/ggml-org/models/resolve/main/tinyllamas/stories260K-be.gguf
           ./bin/llama-completion -m stories260K-be.gguf -p "One day, Lily met a Shoggoth" -n 500 -c 256
 
+  android-arm64:
+    runs-on: ubuntu-latest
+
+    env:
+      NDK_VERSION: "29.0.14206865"
+
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v6
+
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.21
+        with:
+          key: android-arm64
+          evict-old-files: 1d
+          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+
+      - name: Set up JDK
+        uses: actions/setup-java@v5
+        with:
+          java-version: 17
+          distribution: temurin
+
+      - name: Setup Android SDK
+        uses: android-actions/setup-android@40fd30fb8d7440372e1316f5d1809ec01dcd3699 # v4.0.1
+        with:
+          log-accepted-android-sdk-licenses: false
+
+      - name: Install NDK
+        run: |
+          sdkmanager "ndk;${{ env.NDK_VERSION }}"
+          echo "ANDROID_NDK=${ANDROID_SDK_ROOT}/ndk/${{ env.NDK_VERSION }}" >> $GITHUB_ENV
+
+      - name: Build
+        id: cmake_build
+        run: |
+          cmake -B build \
+            -DCMAKE_TOOLCHAIN_FILE=${ANDROID_NDK}/build/cmake/android.toolchain.cmake \
+            -DANDROID_ABI=arm64-v8a \
+            -DANDROID_PLATFORM=android-28 \
+            -DLLAMA_FATAL_WARNINGS=ON \
+            -DGGML_BACKEND_DL=ON \
+            -DGGML_NATIVE=OFF \
+            -DGGML_CPU_ALL_VARIANTS=ON \
+            -DGGML_OPENMP=OFF \
+            -DLLAMA_BUILD_BORINGSSL=ON \
+            -DGGML_RPC=ON
+          time cmake --build build --config Release -j $(nproc)
+
   ubuntu-latest-rpc:
     runs-on: ubuntu-latest
 
@@ -505,186 +555,6 @@ jobs:
             -DGGML_MUSA=ON
           time cmake --build build --config Release -j $(nproc)
 
-  ubuntu-22-sycl:
-    runs-on: ubuntu-22.04
-
-    continue-on-error: true
-
-    steps:
-      - uses: actions/checkout@v6
-
-      - name: add oneAPI to apt
-        shell: bash
-        run: |
-          cd /tmp
-          wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
-          sudo apt-key add GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
-          rm GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
-          sudo add-apt-repository "deb https://apt.repos.intel.com/oneapi all main"
-
-      - name: install oneAPI dpcpp compiler
-        shell: bash
-        run: |
-          sudo apt update
-          sudo apt install intel-oneapi-compiler-dpcpp-cpp libssl-dev
-
-      - name: install oneAPI MKL library
-        shell: bash
-        run: |
-          sudo apt install intel-oneapi-mkl-devel
-
-      - name: Clone
-        id: checkout
-        uses: actions/checkout@v6
-
-      - name: ccache
-        uses: ggml-org/ccache-action@v1.2.21
-        with:
-          key: ubuntu-22-sycl
-          evict-old-files: 1d
-          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
-
-      - name: Build
-        id: cmake_build
-        run: |
-          source /opt/intel/oneapi/setvars.sh
-          cmake -B build \
-            -DGGML_SYCL=ON \
-            -DCMAKE_C_COMPILER=icx \
-            -DCMAKE_CXX_COMPILER=icpx
-          time cmake --build build --config Release -j $(nproc)
-
-  ubuntu-22-sycl-fp16:
-    runs-on: ubuntu-22.04
-
-    continue-on-error: true
-
-    steps:
-      - uses: actions/checkout@v6
-
-      - name: add oneAPI to apt
-        shell: bash
-        run: |
-          cd /tmp
-          wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
-          sudo apt-key add GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
-          rm GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB
-          sudo add-apt-repository "deb https://apt.repos.intel.com/oneapi all main"
-
-      - name: install oneAPI dpcpp compiler
-        shell: bash
-        run: |
-          sudo apt update
-          sudo apt install intel-oneapi-compiler-dpcpp-cpp libssl-dev ninja-build
-
-      - name: install oneAPI MKL library
-        shell: bash
-        run: |
-          sudo apt install intel-oneapi-mkl-devel
-
-      - name: Clone
-        id: checkout
-        uses: actions/checkout@v6
-
-      - name: ccache
-        uses: ggml-org/ccache-action@v1.2.21
-        with:
-          key: ubuntu-22-sycl-fp16
-          evict-old-files: 1d
-          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
-
-      - name: Build
-        id: cmake_build
-        run: |
-          source /opt/intel/oneapi/setvars.sh
-          cmake -B build \
-            -G "Ninja" \
-            -DCMAKE_BUILD_TYPE=Release \
-            -DGGML_SYCL=ON \
-            -DCMAKE_C_COMPILER=icx \
-            -DCMAKE_CXX_COMPILER=icpx \
-            -DGGML_SYCL_F16=ON
-          time cmake --build build --config Release -j $(nproc)
-
-  ubuntu-24-openvino:
-      name: ubuntu-24-openvino-${{ matrix.openvino_device }}
-      strategy:
-        matrix:
-          include:
-            - variant: cpu
-              runner: '"ubuntu-24.04"'
-              openvino_device: "CPU"
-            - variant: gpu
-              runner: '["self-hosted","Linux","X64","Intel"]'
-              openvino_device: "GPU"
-
-      runs-on: ${{ fromJSON(matrix.runner) }}
-
-      env:
-        # Sync versions in build.yml, build-self-hosted.yml, release.yml, build-cache.yml, .devops/openvino.Dockerfile
-        OPENVINO_VERSION_MAJOR: "2026.0"
-        OPENVINO_VERSION_FULL: "2026.0.0.20965.c6d6a13a886"
-
-      steps:
-        - name: Clone
-          id: checkout
-          uses: actions/checkout@v6
-
-        - name: ccache
-          if: runner.environment == 'github-hosted'
-          uses: ggml-org/ccache-action@v1.2.21
-          with:
-            key: ubuntu-24-openvino-${{ matrix.variant }}-no-preset-v1
-            evict-old-files: 1d
-            save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
-
-        - name: Dependencies
-          id: depends
-          run: |
-            sudo apt-get update
-            sudo apt-get install -y build-essential libssl-dev libtbb12 cmake ninja-build python3-pip
-            sudo apt-get install -y ocl-icd-opencl-dev opencl-headers opencl-clhpp-headers intel-opencl-icd
-
-        - name: Use OpenVINO Toolkit Cache
-          if: runner.environment == 'github-hosted'
-          uses: actions/cache@v5
-          id: cache-openvino
-          with:
-            path: ./openvino_toolkit
-            key: openvino-toolkit-v${{ env.OPENVINO_VERSION_FULL }}-${{ runner.os }}
-
-        - name: Setup OpenVINO Toolkit
-          if: steps.cache-openvino.outputs.cache-hit != 'true'
-          uses: ./.github/actions/linux-setup-openvino
-          with:
-            path: ./openvino_toolkit
-            version_major: ${{ env.OPENVINO_VERSION_MAJOR }}
-            version_full: ${{ env.OPENVINO_VERSION_FULL }}
-
-        - name: Install OpenVINO dependencies
-          run: |
-            cd ./openvino_toolkit
-            chmod +x ./install_dependencies/install_openvino_dependencies.sh
-            echo "Y" | sudo -E ./install_dependencies/install_openvino_dependencies.sh
-
-        - name: Build
-          id: cmake_build
-          run: |
-            source ./openvino_toolkit/setupvars.sh
-            cmake -B build/ReleaseOV -G Ninja \
-              -DCMAKE_BUILD_TYPE=Release \
-              -DGGML_OPENVINO=ON
-            time cmake --build build/ReleaseOV --config Release -j $(nproc)
-
-        - name: Test
-          id: cmake_test
-          # TODO: fix and re-enable the `test-llama-archs` test below
-          run: |
-            cd ${{ github.workspace }}
-            if [ "${{ matrix.openvino_device }}" = "GPU" ]; then
-              export GGML_OPENVINO_DEVICE=GPU
-            fi
-            ctest --test-dir build/ReleaseOV -L main -E "test-llama-archs" --verbose --timeout 2000
 
   windows-latest:
     runs-on: windows-2025
@@ -893,39 +763,6 @@ jobs:
           cmake --build build --config Release -j %NINJA_JOBS% -t ggml
           cmake --build build --config Release
 
-  windows-latest-sycl:
-    runs-on: windows-2022
-
-    defaults:
-      run:
-        shell: bash
-
-    env:
-      WINDOWS_BASEKIT_URL: https://registrationcenter-download.intel.com/akdlm/IRC_NAS/24751ead-ddc5-4479-b9e6-f9fe2ff8b9f2/intel-deep-learning-essentials-2025.2.1.25_offline.exe
-      WINDOWS_DPCPP_MKL: intel.oneapi.win.cpp-dpcpp-common:intel.oneapi.win.mkl.devel:intel.oneapi.win.dnnl:intel.oneapi.win.tbb.devel
-      ONEAPI_ROOT: "C:/Program Files (x86)/Intel/oneAPI"
-    steps:
-      - name: Clone
-        id: checkout
-        uses: actions/checkout@v6
-
-      - name: ccache
-        uses: ggml-org/ccache-action@v1.2.21
-        with:
-          key: windows-latest-sycl
-          variant: ccache
-          evict-old-files: 1d
-          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
-
-      - name: Install
-        run:  |
-          scripts/install-oneapi.bat $WINDOWS_BASEKIT_URL $WINDOWS_DPCPP_MKL
-
-      # TODO: add ssl support ; we will also need to modify win-build-sycl.bat to accept user-specified args
-
-      - name: Build
-        id: cmake_build
-        run:  examples/sycl/win-build-sycl.bat
 
   windows-latest-hip:
     runs-on: windows-2022
@@ -1001,22 +838,14 @@ jobs:
     steps:
       - name: Install dependencies
         run: |
-          sudo apt-get update
-
           # Install necessary packages
-          sudo apt-get install -y libatomic1 libtsan2 gcc-14 g++-14 cmake build-essential libssl-dev wget git-lfs
+          sudo apt-get update
+          sudo apt-get install -y libssl-dev
 
           # Set gcc-14 and g++-14 as the default compilers
           sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-14 100
           sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-14 100
 
-          if ! which rustc; then
-            # Install Rust stable version
-            sudo apt-get install -y rustup
-            rustup install stable
-            rustup default stable
-          fi
-
           git lfs install
 
       - name: Check environment
@@ -1032,13 +861,12 @@ jobs:
         id: checkout
         uses: actions/checkout@v6
 
-      # FIXME: Enable when ggml-org/ccache-action works on riscv64
-      # - name: ccache
-      #   uses: ggml-org/ccache-action@v1.2.21
-      #   with:
-      #     key: ubuntu-cpu-riscv64-native
-      #     evict-old-files: 1d
-      #     save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+      - name: ccache
+        uses: ggml-org/ccache-action@afde29e5b5422e5da23cb1f639e8baecadeadfc3 # https://github.com/ggml-org/ccache-action/pull/1
+        with:
+          key: ubuntu-cpu-riscv64-native
+          evict-old-files: 1d
+          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
 
       - name: Build
         id: cmake_build

.github/workflows/python-type-check.yml

@@ -31,7 +31,7 @@ jobs:
         uses: actions/setup-python@v6
         with:
           python-version: "3.11"
-          pip-install: -r requirements/requirements-all.txt ty==0.0.26
+          pip-install: -r requirements/requirements-all.txt ty==0.0.33
       # - name: Type-check with Pyright
       #   uses: jakebailey/pyright-action@v2
       #   with:

.github/workflows/release.yml

@@ -236,6 +236,75 @@ jobs:
           path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-${{ matrix.build }}.tar.gz
           name: llama-bin-ubuntu-vulkan-${{ matrix.build }}.tar.gz
 
+  android-arm64:
+    runs-on: ubuntu-latest
+
+    env:
+      NDK_VERSION: "29.0.14206865"
+
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v6
+        with:
+          fetch-depth: 0
+
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.21
+        with:
+          key: android-arm64
+          evict-old-files: 1d
+
+      - name: Set up JDK
+        uses: actions/setup-java@v5
+        with:
+          java-version: 17
+          distribution: temurin
+
+      - name: Setup Android SDK
+        uses: android-actions/setup-android@40fd30fb8d7440372e1316f5d1809ec01dcd3699 # v4.0.1
+        with:
+          log-accepted-android-sdk-licenses: false
+
+      - name: Install NDK
+        run: |
+          sdkmanager "ndk;${{ env.NDK_VERSION }}"
+          echo "ANDROID_NDK=${ANDROID_SDK_ROOT}/ndk/${{ env.NDK_VERSION }}" >> $GITHUB_ENV
+
+      - name: Build
+        id: cmake_build
+        run: |
+          cmake -B build \
+            -DCMAKE_TOOLCHAIN_FILE=${ANDROID_NDK}/build/cmake/android.toolchain.cmake \
+            -DANDROID_ABI=arm64-v8a \
+            -DANDROID_PLATFORM=android-28 \
+            -DCMAKE_INSTALL_RPATH='$ORIGIN' \
+            -DCMAKE_BUILD_WITH_INSTALL_RPATH=ON \
+            -DGGML_BACKEND_DL=ON \
+            -DGGML_NATIVE=OFF \
+            -DGGML_CPU_ALL_VARIANTS=ON \
+            -DLLAMA_FATAL_WARNINGS=ON \
+            -DGGML_OPENMP=OFF \
+            -DLLAMA_BUILD_BORINGSSL=ON \
+            ${{ env.CMAKE_ARGS }}
+          cmake --build build --config Release -j $(nproc)
+
+      - name: Determine tag name
+        id: tag
+        uses: ./.github/actions/get-tag-name
+
+      - name: Pack artifacts
+        id: pack_artifacts
+        run: |
+          cp LICENSE ./build/bin/
+          tar -czvf llama-${{ steps.tag.outputs.name }}-bin-android-arm64.tar.gz --transform "s,./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
+
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v6
+        with:
+          path: llama-${{ steps.tag.outputs.name }}-bin-android-arm64.tar.gz
+          name: llama-bin-android-arm64.tar.gz
+
   ubuntu-24-openvino:
     runs-on: ubuntu-24.04
 
@@ -529,15 +598,29 @@ jobs:
         shell: bash
 
     env:
-      WINDOWS_BASEKIT_URL: https://registrationcenter-download.intel.com/akdlm/IRC_NAS/24751ead-ddc5-4479-b9e6-f9fe2ff8b9f2/intel-deep-learning-essentials-2025.2.1.25_offline.exe
+      WINDOWS_BASEKIT_URL: https://registrationcenter-download.intel.com/akdlm/IRC_NAS/b60765d1-2b85-4e85-86b6-cb0e9563a699/intel-deep-learning-essentials-2025.3.3.18_offline.exe
       WINDOWS_DPCPP_MKL: intel.oneapi.win.cpp-dpcpp-common:intel.oneapi.win.mkl.devel:intel.oneapi.win.dnnl:intel.oneapi.win.tbb.devel
       ONEAPI_ROOT: "C:/Program Files (x86)/Intel/oneAPI"
+      ONEAPI_INSTALLER_VERSION: "2025.3.3"
 
     steps:
       - name: Clone
         id: checkout
         uses: actions/checkout@v6
 
+      - name: Use oneAPI Installation Cache
+        uses: actions/cache@v5
+        id: cache-sycl
+        with:
+          path: ${{ env.ONEAPI_ROOT }}
+          key: oneAPI-${{ env.ONEAPI_INSTALLER_VERSION }}-${{ runner.os }}
+
+      - name: Download & Install oneAPI
+        shell: bash
+        if: steps.cache-sycl.outputs.cache-hit != 'true'
+        run: |
+          scripts/install-oneapi.bat $WINDOWS_BASEKIT_URL $WINDOWS_DPCPP_MKL
+
       - name: ccache
         uses: ggml-org/ccache-action@v1.2.21
         with:
@@ -545,10 +628,6 @@ jobs:
           variant: ccache
           evict-old-files: 1d
 
-      - name: Install
-        run:  |
-          scripts/install-oneapi.bat $WINDOWS_BASEKIT_URL $WINDOWS_DPCPP_MKL
-
       - name: Build
         id: cmake_build
         shell: cmd
@@ -601,6 +680,82 @@ jobs:
           path: llama-bin-win-sycl-x64.zip
           name: llama-bin-win-sycl-x64.zip
 
+  ubuntu-24-sycl:
+    strategy:
+      matrix:
+        build: [fp32, fp16]
+        include:
+          - build: fp32
+            fp16: OFF
+          - build: fp16
+            fp16: ON
+
+    runs-on: ubuntu-24.04
+
+    env:
+      ONEAPI_ROOT: /opt/intel/oneapi/
+      ONEAPI_INSTALLER_VERSION: "2025.3.3"
+
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v6
+        with:
+          fetch-depth: 0
+
+      - name: Use oneAPI Installation Cache
+        uses: actions/cache@v5
+        id: cache-sycl
+        with:
+          path: ${{ env.ONEAPI_ROOT }}
+          key: oneAPI-${{ env.ONEAPI_INSTALLER_VERSION }}-${{ runner.os }}
+
+      - name: Download & Install oneAPI
+        shell: bash
+        if: steps.cache-sycl.outputs.cache-hit != 'true'
+        run: |
+          cd /tmp
+          wget https://registrationcenter-download.intel.com/akdlm/IRC_NAS/56f7923a-adb8-43f3-8b02-2b60fcac8cab/intel-deep-learning-essentials-2025.3.3.16_offline.sh -O intel-deep-learning-essentials_offline.sh
+          sudo bash intel-deep-learning-essentials_offline.sh -s -a --silent --eula accept
+
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.21
+        with:
+          key: ubuntu-24-sycl-${{ matrix.build }}
+          evict-old-files: 1d
+          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+
+      - name: Build
+        id: cmake_build
+        run: |
+          source /opt/intel/oneapi/setvars.sh
+          cmake -B build \
+            -G "Ninja" \
+            -DCMAKE_BUILD_TYPE=Release \
+            -DGGML_SYCL=ON \
+            -DCMAKE_C_COMPILER=icx \
+            -DCMAKE_CXX_COMPILER=icpx \
+            -DLLAMA_OPENSSL=OFF \
+            -DGGML_NATIVE=OFF \
+            -DGGML_SYCL_F16=${{ matrix.fp16 }}
+          time cmake --build build --config Release -j $(nproc)
+
+      - name: Determine tag name
+        id: tag
+        uses: ./.github/actions/get-tag-name
+
+      - name: Pack artifacts
+        id: pack_artifacts
+        run: |
+          cp LICENSE ./build/bin/
+          tar -czvf llama-${{ steps.tag.outputs.name }}-bin-ubuntu-sycl-${{ matrix.build }}-x64.tar.gz --transform "s,./,llama-${{ steps.tag.outputs.name }}/," -C ./build/bin .
+
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v6
+        with:
+          path: llama-${{ steps.tag.outputs.name }}-bin-ubuntu-sycl-${{ matrix.build }}-x64.tar.gz
+          name: llama-bin-ubuntu-sycl-${{ matrix.build }}-x64.tar.gz
+
   ubuntu-22-rocm:
     runs-on: ubuntu-22.04
 
@@ -618,6 +773,11 @@ jobs:
         with:
           fetch-depth: 0
 
+      - name: Free up disk space
+        uses: ggml-org/free-disk-space@v1.3.1
+        with:
+          tool-cache: true
+
       - name: ccache
         uses: ggml-org/ccache-action@v1.2.21
         with:
@@ -971,6 +1131,8 @@ jobs:
       - ubuntu-cpu
       - ubuntu-vulkan
       - ubuntu-24-openvino
+      - ubuntu-24-sycl
+      - android-arm64
       - macOS-cpu
       - ios-xcode-build
       - openEuler-cann
@@ -1058,6 +1220,11 @@ jobs:
             - [Ubuntu arm64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-vulkan-arm64.tar.gz)
             - [Ubuntu x64 (ROCm 7.2)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-rocm-7.2-x64.tar.gz)
             - [Ubuntu x64 (OpenVINO)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-openvino-${{ needs.ubuntu-24-openvino.outputs.openvino_version }}-x64.tar.gz)
+            - [Ubuntu x64 (SYCL FP32)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-sycl-fp32-x64.tar.gz)
+            - [Ubuntu x64 (SYCL FP16)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-ubuntu-sycl-fp16-x64.tar.gz)
+
+            **Android:**
+            - [Android arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-android-arm64.tar.gz)
 
             **Windows:**
             - [Windows x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/${{ steps.tag.outputs.name }}/llama-${{ steps.tag.outputs.name }}-bin-win-cpu-x64.zip)

.gitignore

@@ -34,7 +34,6 @@
 /.vscode/
 /nppBackup
 
-
 # Coverage
 
 /gcovr-report/
@@ -74,6 +73,7 @@
 !/models/templates
 
 # Zig
+
 /zig-out/
 /zig-cache/
 
@@ -93,6 +93,7 @@
 !/examples/sycl/*.sh
 
 # Server Web UI temporary files
+
 /tools/server/webui/node_modules
 /tools/server/webui/dist
 # we no longer use gz for index.html
@@ -106,9 +107,11 @@ __pycache__/
 poetry.toml
 
 # Nix
+
 /result
 
 # Test binaries
+
 /tests/test-backend-ops
 /tests/test-double-float
 /tests/test-grad0
@@ -124,6 +127,7 @@ poetry.toml
 /tests/test-tokenizer-1-spm
 
 # Scripts
+
 !/scripts/install-oneapi.bat
 
 # Generated by scripts
@@ -132,16 +136,24 @@ poetry.toml
 /wikitext-2-raw/
 
 # Test models for lora adapters
+
 /lora-tests
 
 # Local scripts
+
 /run-vim.sh
 /run-chat.sh
 /run-spec.sh
 /.ccache/
 
 # IDE
+
 /*.code-workspace
 /.windsurf/
 # emscripten
 a.out.*
+
+# AGENTS
+
+AGENTS.local.md
+.pi/SYSTEM.md

.pi/gg/SYSTEM.md

@@ -0,0 +1,33 @@
+You are a coding agent. Here are some very important rules that you must follow:
+
+General:
+- By very precise and concise when writing code, comments, explanations, etc.
+- PR and commit titles format: `<module> : <title>`. Lookup recents for examples
+- Don't try to build or run the code unless you are explicitly asked to do so
+
+Coding:
+- When in doubt, always refer to the CONTRIBUTING.md file of the project
+- When referencing issues or PRs in comments, use the format:
+  - C/C++ code: `// ref: <url>`
+  - Other (CMake, etc.): `# ref: <url>`
+
+Pull requests (PRs):
+- New branch names are prefixed with "gg/"
+- Before opening a pull request, ask the user to confirm the description
+- When creating a pull request, look for the repository's PR template and follow it
+- For the AI usage disclosure section, write "YES. llama.cpp + pi"
+- Always create the pull requests in draft mode
+
+Commits:
+- On every commit that you make, include a "Assisted-by: llama.cpp:local pi" tag
+- Do not explicitly set the git author in commits - rely on the default git config
+
+Resources (read on demand):
+- [CONTRIBUTING.md](CONTRIBUTING.md)
+- [Build documentation](docs/build.md)
+- [Server usage documentation](tools/server/README.md)
+- [Server development documentation](tools/server/README-dev.md)
+- [PEG parser](docs/development/parsing.md)
+- [Auto parser](docs/autoparser.md)
+- [Jinja engine](common/jinja/README.md)
+- [PR template](.github/pull_request_template.md)

CMakeLists.txt

@@ -225,7 +225,7 @@ foreach(FILE_PATH ${EXTRA_LICENSES})
 endforeach()
 
 if (LLAMA_BUILD_COMMON)
-    license_generate(common)
+    license_generate(llama-common)
 endif()
 
 #
@@ -249,6 +249,10 @@ set_target_properties(llama
 
 install(TARGETS llama LIBRARY PUBLIC_HEADER)
 
+if (LLAMA_BUILD_COMMON)
+    install(TARGETS llama-common LIBRARY)
+endif()
+
 configure_package_config_file(
         ${CMAKE_CURRENT_SOURCE_DIR}/cmake/llama-config.cmake.in
         ${CMAKE_CURRENT_BINARY_DIR}/llama-config.cmake

CODEOWNERS

@@ -1,5 +1,21 @@
 # collaborators can optionally add themselves here to indicate their availability for reviewing related PRs
-# multiplie collaborators per item can be specified
+# multiple collaborators per item can be specified
+#
+# ggml-org/ci               : CISC, danbev, ggerganov, netrunnereve, ngxson, taronaeo
+# ggml-org/ggml-cann        : hipudding
+# ggml-org/ggml-cuda        : JohannesGaessler, am17an, IMbackK, ORippler
+# ggml-org/ggml-hexagon     : lhez, max-krasnyansky
+# ggml-org/ggml-metal       : ggerganov
+# ggml-org/ggml-opencl      : lhez, max-krasnyansky
+# ggml-org/ggml-rpc         : rgerganov
+# ggml-org/ggml-sycl        : arthw
+# ggml-org/ggml-vulkan      : 0cc4m, jeffbolznv
+# ggml-org/ggml-webgpu      : reeselevine
+# ggml-org/ggml-zdnn        : taronaeo
+# ggml-org/llama-common     : ggerganov, aldehir, angt, danbev, ngxson, pwilkin
+# ggml-org/llama-mtmd       : ngxson
+# ggml-org/llama-server     : ggerganov, ngxson, allozaur, angt, ServeurpersoCom
+# ggml-org/llama-webui      : allozaur
 
 /.devops/*.Dockerfile                   @ngxson
 /.github/actions/                       @ggml-org/ci
@@ -7,6 +23,7 @@
 /ci/                                    @ggerganov
 /cmake/                                 @ggerganov
 /common/                                @ggml-org/llama-common
+/common/fit.*                           @JohannesGaessler
 /common/jinja/                          @CISC
 /common/ngram-map.*                     @srogmann
 /convert_*.py                           @CISC
@@ -36,28 +53,29 @@
 /examples/speculative/                  @ggerganov
 /ggml/cmake/                            @ggerganov
 /ggml/include/                          @ggerganov
+/ggml/src/ggml-backend-meta.cpp         @JohannesGaessler
 /ggml/src/ggml-cann/                    @ggml-org/ggml-cann
 /ggml/src/ggml-common.h                 @ggerganov
 /ggml/src/ggml-cpu/                     @ggerganov
 /ggml/src/ggml-cpu/spacemit/            @alex-spacemit
 /ggml/src/ggml-cuda/                    @ggml-org/ggml-cuda
-/ggml/src/ggml-cuda/fattn-wmma*         @IMbackK
-/ggml/src/ggml-hip/                     @IMbackK
 /ggml/src/ggml-cuda/vendors/hip.h       @IMbackK
+/ggml/src/ggml-cuda/fattn-wmma*         @IMbackK
+/ggml/src/ggml-hexagon/                 @ggml-org/ggml-hexagon
+/ggml/src/ggml-hip/                     @IMbackK
 /ggml/src/ggml-impl.h                   @ggerganov
 /ggml/src/ggml-metal/                   @ggml-org/ggml-metal
 /ggml/src/ggml-opencl/                  @ggml-org/ggml-opencl
-/ggml/src/ggml-hexagon/                 @ggml-org/ggml-hexagon
+/ggml/src/ggml-openvino/                @cavusmustafa @wine99
 /ggml/src/ggml-opt.cpp                  @JohannesGaessler
 /ggml/src/ggml-quants.*                 @ggerganov
 /ggml/src/ggml-rpc/                     @ggml-org/ggml-rpc
 /ggml/src/ggml-sycl/                    @ggml-org/ggml-sycl
 /ggml/src/ggml-threading.*              @ggerganov
-/ggml/src/ggml-vulkan/                  @ggml-org/ggml-vulkan
 /ggml/src/ggml-virtgpu/                 @kpouget
+/ggml/src/ggml-vulkan/                  @ggml-org/ggml-vulkan
 /ggml/src/ggml-webgpu/                  @ggml-org/ggml-webgpu
 /ggml/src/ggml-zdnn/                    @ggml-org/ggml-zdnn @Andreas-Krebbel @AlekseiNikiforovIBM
-/ggml/src/ggml-openvino/                @cavusmustafa @wine99
 /ggml/src/ggml.c                        @ggerganov
 /ggml/src/ggml.cpp                      @ggerganov
 /ggml/src/gguf.cpp                      @JohannesGaessler @Green-Sky

common/CMakeLists.txt

@@ -1,9 +1,11 @@
-# common
-
 find_package(Threads REQUIRED)
 
 llama_add_compile_flags()
 
+#
+# llama-common-base
+#
+
 # Build info header
 
 if(EXISTS "${PROJECT_SOURCE_DIR}/.git")
@@ -33,17 +35,25 @@ endif()
 
 set(TEMPLATE_FILE "${CMAKE_CURRENT_SOURCE_DIR}/build-info.cpp.in")
 set(OUTPUT_FILE   "${CMAKE_CURRENT_BINARY_DIR}/build-info.cpp")
+
 configure_file(${TEMPLATE_FILE} ${OUTPUT_FILE})
 
-set(TARGET build_info)
-add_library(${TARGET} OBJECT ${OUTPUT_FILE})
+set(TARGET llama-common-base)
+add_library(${TARGET} STATIC ${OUTPUT_FILE})
+
+target_include_directories(${TARGET} PUBLIC .)
+
 if (BUILD_SHARED_LIBS)
     set_target_properties(${TARGET} PROPERTIES POSITION_INDEPENDENT_CODE ON)
 endif()
 
-set(TARGET common)
+#
+# llama-common
+#
 
-add_library(${TARGET} STATIC
+set(TARGET llama-common)
+
+add_library(${TARGET}
     arg.cpp
     arg.h
     base64.hpp
@@ -63,6 +73,8 @@ add_library(${TARGET} STATIC
     debug.h
     download.cpp
     download.h
+    fit.cpp
+    fit.h
     hf-cache.cpp
     hf-cache.h
     http.h
@@ -106,17 +118,24 @@ add_library(${TARGET} STATIC
     jinja/caps.h
     )
 
+set_target_properties(${TARGET} PROPERTIES
+    VERSION ${LLAMA_INSTALL_VERSION}
+    SOVERSION 0
+    MACHO_CURRENT_VERSION 0 # keep macOS linker from seeing oversized version number
+)
+
 target_include_directories(${TARGET} PUBLIC . ../vendor)
 target_compile_features   (${TARGET} PUBLIC cxx_std_17)
 
 if (BUILD_SHARED_LIBS)
     set_target_properties(${TARGET} PROPERTIES POSITION_INDEPENDENT_CODE ON)
+
+    # TODO: make fine-grained exports in the future
+    set_target_properties(${TARGET} PROPERTIES WINDOWS_EXPORT_ALL_SYMBOLS ON)
 endif()
 
-target_link_libraries(${TARGET} PRIVATE
-    build_info
-    cpp-httplib
-)
+target_link_libraries(${TARGET} PUBLIC  llama-common-base)
+target_link_libraries(${TARGET} PRIVATE cpp-httplib)
 
 if (LLAMA_LLGUIDANCE)
     include(ExternalProject)

common/arg.h

@@ -25,7 +25,8 @@ struct common_arg {
     const char * value_hint_2 = nullptr; // for second arg value
     const char * env          = nullptr;
     std::string help;
-    bool is_sparam = false; // is current arg a sampling param?
+    bool is_sampling = false; // is current arg a sampling param?
+    bool is_spec = false; // is current arg a speculative decoding param?
     bool is_preset_only = false; // is current arg preset-only (not treated as CLI arg)
     void (*handler_void)   (common_params & params) = nullptr;
     void (*handler_string) (common_params & params, const std::string &) = nullptr;
@@ -74,7 +75,8 @@ struct common_arg {
     common_arg & set_examples(std::initializer_list<enum llama_example> examples);
     common_arg & set_excludes(std::initializer_list<enum llama_example> excludes);
     common_arg & set_env(const char * env);
-    common_arg & set_sparam();
+    common_arg & set_sampling();
+    common_arg & set_spec();
     common_arg & set_preset_only();
     bool in_example(enum llama_example ex);
     bool is_exclude(enum llama_example ex);

common/build-info.cpp.in

@@ -1,4 +1,35 @@
+#include "build-info.h"
+
+#include <cstdio>
+#include <string>
+
 int LLAMA_BUILD_NUMBER = @LLAMA_BUILD_NUMBER@;
-char const *LLAMA_COMMIT = "@LLAMA_BUILD_COMMIT@";
-char const *LLAMA_COMPILER = "@BUILD_COMPILER@";
-char const *LLAMA_BUILD_TARGET = "@BUILD_TARGET@";
+char const * LLAMA_COMMIT = "@LLAMA_BUILD_COMMIT@";
+char const * LLAMA_COMPILER = "@BUILD_COMPILER@";
+char const * LLAMA_BUILD_TARGET = "@BUILD_TARGET@";
+
+int llama_build_number(void) {
+    return LLAMA_BUILD_NUMBER;
+}
+
+const char * llama_commit(void) {
+    return LLAMA_COMMIT;
+}
+
+const char * llama_compiler(void) {
+    return LLAMA_COMPILER;
+}
+
+const char * llama_build_target(void) {
+    return LLAMA_BUILD_TARGET;
+}
+
+const char * llama_build_info(void) {
+    static std::string s = "b" + std::to_string(LLAMA_BUILD_NUMBER) + "-" + LLAMA_COMMIT;
+    return s.c_str();
+}
+
+void llama_print_build_info(void) {
+    fprintf(stderr, "%s: build = %d (%s)\n",      __func__, llama_build_number(), llama_commit());
+    fprintf(stderr, "%s: built with %s for %s\n", __func__, llama_compiler(), llama_build_target());
+}

common/build-info.h

@@ -0,0 +1,11 @@
+#pragma once
+
+int llama_build_number(void);
+
+const char * llama_commit(void);
+const char * llama_compiler(void);
+
+const char * llama_build_target(void);
+const char * llama_build_info(void);
+
+void llama_print_build_info(void);

common/chat-auto-parser-generator.cpp

@@ -443,14 +443,14 @@ common_peg_parser analyze_tools::build_tool_parser_tag_tagged(parser_build_conte
     if (!format.per_call_start.empty()) {
         auto wrapped_call = format.per_call_start + p.space() + tool_choice + p.space() + format.per_call_end;
         if (inputs.parallel_tool_calls) {
-            tool_calls = p.trigger_rule("tool-call", wrapped_call + p.zero_or_more(p.space() + wrapped_call));
+            tool_calls = p.trigger_rule("tool-call", wrapped_call + p.zero_or_more(p.space() + wrapped_call) + p.space());
         } else {
-            tool_calls = p.trigger_rule("tool-call", wrapped_call);
+            tool_calls = p.trigger_rule("tool-call", wrapped_call + p.space());
         }
         if (!format.section_start.empty()) {
             tool_calls = p.trigger_rule("tool-calls",
                                         p.literal(format.section_start) + p.space() + tool_calls + p.space() +
-                                            (format.section_end.empty() ? p.end() : p.literal(format.section_end)));
+                                            (format.section_end.empty() ? p.end() : p.literal(format.section_end) + p.space()));
         }
     } else {
         std::string separator = ", ";  // Default

common/chat-diff-analyzer.cpp

@@ -296,7 +296,7 @@ void analyze_reasoning::compare_reasoning_presence() {
             return p.literal(reasoning_content) + p.space() + p.optional(p.tag("post", (p.marker() + p.space())) + p.rest());
         });
         auto parser_wrapped = build_tagged_peg_parser([&](common_peg_parser_builder &p) {
-            return p.tag("pre", p.marker() + p.space()) + p.literal(reasoning_content) + p.space() + p.tag("post", (p.marker() + p.space())) + p.rest();
+            return p.tag("pre", p.marker() + p.space()) + p.literal(reasoning_content) + p.tag("post", (p.space() + p.marker() + p.space())) + p.rest();
         });
         // try the more aggressive parse first, if it fails, fall back to the delimiter one
         auto result = parser_wrapped.parse_anywhere_and_extract(comparison->output_B);
@@ -306,11 +306,11 @@ void analyze_reasoning::compare_reasoning_presence() {
         if (result.result.success()) {
             if (!result.tags["pre"].empty() && !result.tags["post"].empty()) {
                 mode = reasoning_mode::TAG_BASED;
-                start = trim_leading_whitespace(result.tags["pre"]);
-                end   = trim_trailing_whitespace(result.tags["post"]);
+                start = result.tags["pre"];
+                end   = result.tags["post"];
             } else if (!result.tags["post"].empty()) {
                 mode = reasoning_mode::TAG_BASED;
-                end = trim_trailing_whitespace(result.tags["post"]);
+                end = result.tags["post"];
             }
         }
     }

common/chat.cpp

@@ -397,6 +397,25 @@ json common_chat_msgs_to_json_oaicompat(const std::vector<common_chat_msg> & msg
     return render_message_to_json(msgs, c);
 }
 
+json common_chat_tools_to_json_oaicompat(const std::vector<common_chat_tool> & tools) {
+    if (tools.empty()) {
+        return json();
+    }
+
+    auto result = json::array();
+    for (const auto & tool : tools) {
+        result.push_back({
+            { "type",     "function" },
+            { "function", {
+                { "name", tool.name },
+                { "description", tool.description },
+                { "parameters", json::parse(tool.parameters) },
+            }},
+        });
+    }
+    return result;
+}
+
 std::vector<common_chat_tool> common_chat_tools_parse_oaicompat(const json & tools) {
     std::vector<common_chat_tool> result;
 
@@ -432,56 +451,6 @@ std::vector<common_chat_tool> common_chat_tools_parse_oaicompat(const json & too
     return result;
 }
 
-json common_chat_tools_to_json_oaicompat(const std::vector<common_chat_tool> & tools) {
-    if (tools.empty()) {
-        return json();
-    }
-
-    auto result = json::array();
-    for (const auto & tool : tools) {
-        result.push_back({
-            { "type",     "function" },
-            { "function",
-             {
-                  { "name", tool.name },
-                  { "description", tool.description },
-                  { "parameters", json::parse(tool.parameters) },
-              }                      },
-        });
-    }
-    return result;
-}
-
-json common_chat_msg_diff_to_json_oaicompat(const common_chat_msg_diff & diff) {
-    json delta = json::object();
-    if (!diff.reasoning_content_delta.empty()) {
-        delta["reasoning_content"] = diff.reasoning_content_delta;
-    }
-    if (!diff.content_delta.empty()) {
-        delta["content"] = diff.content_delta;
-    }
-    if (diff.tool_call_index != std::string::npos) {
-        json tool_call;
-        tool_call["index"] = diff.tool_call_index;
-        if (!diff.tool_call_delta.id.empty()) {
-            tool_call["id"]   = diff.tool_call_delta.id;
-            tool_call["type"] = "function";
-        }
-        if (!diff.tool_call_delta.name.empty() || !diff.tool_call_delta.arguments.empty()) {
-            json function = json::object();
-            if (!diff.tool_call_delta.name.empty()) {
-                function["name"] = diff.tool_call_delta.name;
-            }
-            if (!diff.tool_call_delta.arguments.empty()) {
-                function["arguments"] = diff.tool_call_delta.arguments;
-            }
-            tool_call["function"] = function;
-        }
-        delta["tool_calls"] = json::array({ tool_call });
-    }
-    return delta;
-}
-
 bool common_chat_verify_template(const std::string & tmpl, bool use_jinja) {
     if (use_jinja) {
         try {
@@ -575,6 +544,26 @@ bool common_chat_templates_was_explicit(const struct common_chat_templates * tmp
     return tmpls->has_explicit_template;
 }
 
+// LFM2 format detection: template uses <|tool_list_start|>[...]<|tool_list_end|> around the tool list
+// and <|tool_call_start|>[...]<|tool_call_end|> around each tool call
+static bool is_lfm2_template(const std::string & src) {
+    return src.find("<|tool_list_start|>") != std::string::npos &&
+           src.find("<|tool_list_end|>")   != std::string::npos;
+}
+
+common_chat_prompt_preset common_chat_get_asr_prompt(const common_chat_templates * chat_templates) {
+    common_chat_prompt_preset asr_preset;
+    asr_preset.system = "";
+    asr_preset.user   = "Transcribe audio to text";
+
+    if (chat_templates && chat_templates->template_default && is_lfm2_template(chat_templates->template_default->source())) {
+        asr_preset.system = "Perform ASR.";
+        asr_preset.user   = "";
+    }
+
+    return asr_preset;
+}
+
 std::string common_chat_templates_source(const struct common_chat_templates * tmpls, const std::string & variant) {
     if (!variant.empty()) {
         if (variant == "tool_use") {
@@ -2084,10 +2073,7 @@ std::optional<common_chat_params> common_chat_try_specialized_template(
         return common_chat_params_init_kimi_k2(tmpl, params);
     }
 
-    // LFM2 format detection: template uses <|tool_list_start|>[...]<|tool_list_end|> around the tool list
-    // and <|tool_call_start|>[...]<|tool_call_end|> around each tool call
-    if (src.find("<|tool_list_start|>") != std::string::npos &&
-        src.find("<|tool_list_end|>") != std::string::npos) {
+    if (is_lfm2_template(src)) {
         LOG_DBG("Using specialized template: LFM2\n");
         return common_chat_params_init_lfm2(tmpl, params);
     }
@@ -2334,7 +2320,7 @@ common_chat_msg common_chat_peg_parse(const common_peg_arena &          src_pars
         ? input
         : params.generation_prompt + input;
 
-    LOG_DBG("Parsing PEG input with format %s: %s\n", common_chat_format_name(params.format), effective_input.c_str());
+    //LOG_DBG("Parsing PEG input with format %s: %s\n", common_chat_format_name(params.format), effective_input.c_str());
 
     common_peg_parse_flags flags = COMMON_PEG_PARSE_FLAG_LENIENT;
     if (params.debug) {
@@ -2396,4 +2382,3 @@ std::map<std::string, bool> common_chat_templates_get_caps(const common_chat_tem
     GGML_ASSERT(chat_templates->template_default != nullptr);
     return chat_templates->template_default->caps.to_map();
 }
-

common/chat.h

@@ -256,14 +256,13 @@ bool common_chat_templates_support_enable_thinking(const common_chat_templates *
 // Parses a JSON array of messages in OpenAI's chat completion API format.
 std::vector<common_chat_msg> common_chat_msgs_parse_oaicompat(const nlohmann::ordered_json & messages);
 
+std::vector<common_chat_tool> common_chat_tools_parse_oaicompat(const nlohmann::ordered_json & tools);
+
 // DEPRECATED: only used in tests
 nlohmann::ordered_json common_chat_msgs_to_json_oaicompat(const std::vector<common_chat_msg> & msgs, bool concat_typed_text = false);
 
-std::vector<common_chat_tool> common_chat_tools_parse_oaicompat(const nlohmann::ordered_json & tools);
 nlohmann::ordered_json common_chat_tools_to_json_oaicompat(const std::vector<common_chat_tool> & tools);
 
-nlohmann::ordered_json common_chat_msg_diff_to_json_oaicompat(const common_chat_msg_diff & diff);
-
 // get template caps, useful for reporting to server /props endpoint
 std::map<std::string, bool> common_chat_templates_get_caps(const common_chat_templates * chat_templates);
 
@@ -275,3 +274,11 @@ std::optional<common_chat_params> common_chat_try_specialized_template(
         const common_chat_template &          tmpl,
         const std::string &                   src,
         autoparser::generation_params & params);
+
+// specialized per-task preset
+struct common_chat_prompt_preset {
+    std::string system;
+    std::string user;
+};
+
+common_chat_prompt_preset common_chat_get_asr_prompt(const common_chat_templates * chat_templates);

common/common.cpp

@@ -1,7 +1,9 @@
 #include "ggml.h"
 #include "gguf.h"
 
+#include "build-info.h"
 #include "common.h"
+#include "fit.h"
 #include "log.h"
 #include "llama.h"
 #include "sampling.h"
@@ -68,7 +70,7 @@ common_time_meas::~common_time_meas() {
 // CPU utils
 //
 
-int32_t cpu_get_num_physical_cores() {
+int32_t common_cpu_get_num_physical_cores() {
 #ifdef __linux__
     // enumerate the set of thread siblings, num entries is num cores
     std::unordered_set<std::string> siblings;
@@ -183,11 +185,11 @@ static int cpu_count_math_cpus(int n_cpu) {
 /**
  * Returns number of CPUs on system that are useful for math.
  */
-int32_t cpu_get_num_math() {
+int32_t common_cpu_get_num_math() {
 #if defined(__x86_64__) && defined(__linux__) && !defined(__ANDROID__)
     int n_cpu = sysconf(_SC_NPROCESSORS_ONLN);
     if (n_cpu < 1) {
-        return cpu_get_num_physical_cores();
+        return common_cpu_get_num_physical_cores();
     }
     if (is_hybrid_cpu()) {
         cpu_set_t affinity;
@@ -200,7 +202,7 @@ int32_t cpu_get_num_math() {
         }
     }
 #endif
-    return cpu_get_num_physical_cores();
+    return common_cpu_get_num_physical_cores();
 }
 
 // Helper for setting process priority
@@ -261,7 +263,7 @@ bool set_process_priority(enum ggml_sched_priority prio) {
 //
 
 
-void postprocess_cpu_params(cpu_params& cpuparams, const cpu_params* role_model) {
+void postprocess_cpu_params(common_cpu_params & cpuparams, const common_cpu_params * role_model) {
     int32_t n_set = 0;
 
     if (cpuparams.n_threads < 0) {
@@ -269,7 +271,7 @@ void postprocess_cpu_params(cpu_params& cpuparams, const cpu_params* role_model)
         if (role_model != nullptr) {
             cpuparams = *role_model;
         } else {
-            cpuparams.n_threads = cpu_get_num_math();
+            cpuparams.n_threads = common_cpu_get_num_math();
         }
     }
 
@@ -372,7 +374,7 @@ void common_init() {
     const char * build_type = " (debug)";
 #endif
 
-    LOG_DBG("build: %d (%s) with %s for %s%s\n", LLAMA_BUILD_NUMBER, LLAMA_COMMIT, LLAMA_COMPILER, LLAMA_BUILD_TARGET, build_type);
+    LOG_DBG("build: %d (%s) with %s for %s%s\n", llama_build_number(), llama_commit(), llama_compiler(), llama_build_target(), build_type);
 }
 
 std::string common_params_get_system_info(const common_params & params) {
@@ -1146,7 +1148,7 @@ common_init_result::common_init_result(common_params & params) :
 
     if (params.fit_params) {
         LOG_INF("%s: fitting params to device memory, for bugs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on\n", __func__);
-        llama_params_fit(params.model.path.c_str(), &mparams, &cparams,
+        common_fit_params(params.model.path.c_str(), &mparams, &cparams,
             params.tensor_split,
             params.tensor_buft_overrides.data(),
             params.fit_params_target.data(),
@@ -1381,7 +1383,7 @@ common_init_result_ptr common_init_from_params(common_params & params) {
 
 common_init_result::~common_init_result() = default;
 
-std::string get_model_endpoint() {
+std::string common_get_model_endpoint() {
     const char * model_endpoint_env = getenv("MODEL_ENDPOINT");
     // We still respect the use of environment-variable "HF_ENDPOINT" for backward-compatibility.
     const char * hf_endpoint_env = getenv("HF_ENDPOINT");
@@ -1396,6 +1398,42 @@ std::string get_model_endpoint() {
     return model_endpoint;
 }
 
+common_context_seq_rm_type common_context_can_seq_rm(llama_context * ctx) {
+    auto * mem = llama_get_memory(ctx);
+    if (mem == nullptr) {
+        return COMMON_CONTEXT_SEQ_RM_TYPE_NO;
+    }
+
+    common_context_seq_rm_type res = COMMON_CONTEXT_SEQ_RM_TYPE_PART;
+
+    llama_memory_clear(mem, true);
+
+    // eval 2 tokens to check if the context is compatible
+    std::vector<llama_token> tmp;
+    tmp.push_back(0);
+    tmp.push_back(0);
+
+    int ret = llama_decode(ctx, llama_batch_get_one(tmp.data(), tmp.size()));
+    if (ret != 0) {
+        LOG_ERR("%s: llama_decode() failed: %d\n", __func__, ret);
+        res = COMMON_CONTEXT_SEQ_RM_TYPE_NO;
+        goto done;
+    }
+
+    // try to remove the last tokens
+    if (!llama_memory_seq_rm(mem, 0, 1, -1)) {
+        LOG_WRN("%s: the target context does not support partial sequence removal\n", __func__);
+        res = COMMON_CONTEXT_SEQ_RM_TYPE_FULL;
+        goto done;
+    }
+
+done:
+    llama_memory_clear(mem, true);
+    llama_synchronize(ctx);
+
+    return res;
+}
+
 void common_set_adapter_lora(struct llama_context * ctx, std::vector<common_adapter_lora_info> & lora) {
     std::vector<llama_adapter_lora *> loras;
     std::vector<float> scales;
@@ -1483,7 +1521,7 @@ struct llama_context_params common_context_params_to_llama(const common_params &
     return cparams;
 }
 
-struct ggml_threadpool_params ggml_threadpool_params_from_cpu_params(const cpu_params & params) {
+struct ggml_threadpool_params ggml_threadpool_params_from_cpu_params(const common_cpu_params & params) {
     struct ggml_threadpool_params tpp;
 
     ggml_threadpool_params_init(&tpp, params.n_threads); // setup the defaults

common/common.h

@@ -2,15 +2,15 @@
 
 #pragma once
 
+#include "llama-cpp.h"
+
 #include "ggml-opt.h"
 #include "ggml.h"
-#include "llama-cpp.h"
 
 #include <set>
 #include <sstream>
 #include <string>
 #include <string_view>
-#include <variant>
 #include <vector>
 #include <map>
 
@@ -27,11 +27,6 @@
 #define die(msg)          do { fputs("error: " msg "\n", stderr);                exit(1); } while (0)
 #define die_fmt(fmt, ...) do { fprintf(stderr, "error: " fmt "\n", __VA_ARGS__); exit(1); } while (0)
 
-#define print_build_info() do {                                                                     \
-    fprintf(stderr, "%s: build = %d (%s)\n",      __func__, LLAMA_BUILD_NUMBER, LLAMA_COMMIT);      \
-    fprintf(stderr, "%s: built with %s for %s\n", __func__, LLAMA_COMPILER, LLAMA_BUILD_TARGET);    \
-} while(0)
-
 struct common_time_meas {
     common_time_meas(int64_t & t_acc, bool disable = false);
     ~common_time_meas();
@@ -53,21 +48,13 @@ struct common_adapter_lora_info {
 
 using llama_tokens = std::vector<llama_token>;
 
-// build info
-extern int LLAMA_BUILD_NUMBER;
-extern const char * LLAMA_COMMIT;
-extern const char * LLAMA_COMPILER;
-extern const char * LLAMA_BUILD_TARGET;
-
-const static std::string build_info("b" + std::to_string(LLAMA_BUILD_NUMBER) + "-" + LLAMA_COMMIT);
-
 struct common_control_vector_load_info;
 
 //
 // CPU utils
 //
 
-struct cpu_params {
+struct common_cpu_params {
     int      n_threads                   = -1;
     bool     cpumask[GGML_MAX_N_THREADS] = {false}; // CPU affinity mask.
     bool     mask_valid                  = false;   // Default: any CPU
@@ -76,8 +63,8 @@ struct cpu_params {
     uint32_t poll                        = 50;      // Polling (busywait) level (0 - no polling, 100 - mostly polling)
 };
 
-int32_t cpu_get_num_physical_cores();
-int32_t cpu_get_num_math();
+int32_t common_cpu_get_num_physical_cores();
+int32_t common_cpu_get_num_math();
 
 //
 // Common params
@@ -287,6 +274,7 @@ struct common_params_sampling {
     std::vector<llama_token> reasoning_budget_start;           // start tag token sequence
     std::vector<llama_token> reasoning_budget_end;             // end tag token sequence
     std::vector<llama_token> reasoning_budget_forced;          // forced sequence (message + end tag)
+    std::string              reasoning_budget_message;         // message injected before end tag when budget exhausted
 
     bool backend_sampling = false;
 
@@ -309,34 +297,19 @@ struct common_params_model {
 
 struct common_ngram_mod;
 
-struct common_params_speculative {
-    common_speculative_type type = COMMON_SPECULATIVE_TYPE_NONE; // type of speculative decoding
+// draft-model-based speculative decoding parameters
+struct common_params_speculative_draft {
+    int32_t n_max = 16; // maximum number of tokens to draft during speculative decoding
+    int32_t n_min = 0;  // minimum number of draft tokens to use for speculative decoding
 
-    // general-purpose speculative decoding parameters
+    float p_split = 0.1f;  // speculative decoding split probability
+    float p_min   = 0.75f; // minimum speculative decoding probability (greedy)
 
-    int32_t n_max   = 16; // maximum number of tokens to draft during speculative decoding
-    int32_t n_min   = 0; // minimum number of draft tokens to use for speculative decoding
-    float   p_split = 0.1f; // speculative decoding split probability
-    float   p_min   = 0.75f; // minimum speculative decoding probability (greedy)
+    common_params_model mparams;
 
-    // ngram-based speculative decoding
+    llama_model * model = nullptr; // a llama_model that can be shared by multiple speculative contexts
 
-    uint16_t ngram_size_n     = 12; // ngram size for lookup
-    uint16_t ngram_size_m     = 48; // mgram size for speculative tokens
-    uint16_t ngram_min_hits   =  1; // minimum hits at ngram/mgram lookup for mgram to be proposed
-
-    std::shared_ptr<common_ngram_mod> ngram_mod;
-
-    std::string lookup_cache_static;  // path of static ngram cache file for lookup decoding           // NOLINT
-    std::string lookup_cache_dynamic; // path of dynamic ngram cache file for lookup decoding          // NOLINT
-
-    // draft-model speculative decoding
-
-    struct common_params_model mparams_dft;
-
-    llama_model * model_dft = nullptr; // a llama_model that can be shared by multiple speculative contexts
-
-    llama_context_params cparams_dft; // these are the parameters for the draft llama_context
+    llama_context_params cparams; // these are the parameters for the draft llama_context
 
     int32_t n_ctx        = 0;  // draft context size
     int32_t n_gpu_layers = -1; // number of layers to store in VRAM for the draft model (-1 - use default)
@@ -344,25 +317,60 @@ struct common_params_speculative {
     ggml_type cache_type_k = GGML_TYPE_F16; // KV cache data type for the K
     ggml_type cache_type_v = GGML_TYPE_F16; // KV cache data type for the V
 
-    struct cpu_params cpuparams;
-    struct cpu_params cpuparams_batch;
+    common_cpu_params cpuparams;
+    common_cpu_params cpuparams_batch;
 
     std::vector<ggml_backend_dev_t> devices; // devices to use for offloading
 
     std::vector<std::pair<std::string, std::string>> replacements; // main to speculative model replacements
     std::vector<llama_model_tensor_buft_override> tensor_buft_overrides;
+};
+
+struct common_params_speculative_ngram_mod {
+    int32_t n_match = 24;
+
+    int32_t n_max = 64;
+    int32_t n_min = 48;
+
+    // shared instance of the ngram container for all speculative decoding contexts
+    std::shared_ptr<common_ngram_mod> obj;
+};
+
+struct common_params_speculative_ngram_map {
+    uint16_t size_n   = 12; // ngram size for lookup
+    uint16_t size_m   = 48; // mgram size for speculative tokens
+    uint16_t min_hits = 1;  // minimum hits at ngram/mgram lookup for mgram to be proposed
+};
+
+struct common_params_speculative_ngram_cache {
+    std::string lookup_cache_static;  // path of static ngram cache file for lookup decoding
+    std::string lookup_cache_dynamic; // path of dynamic ngram cache file for lookup decoding
+};
+
+struct common_params_speculative {
+    // TODO: become a vector in order to support "chains of speculators"
+    common_speculative_type type = COMMON_SPECULATIVE_TYPE_NONE;
+
+    common_params_speculative_draft draft;
+
+    common_params_speculative_ngram_mod ngram_mod;
+    common_params_speculative_ngram_map ngram_simple;
+    common_params_speculative_ngram_map ngram_map_k;
+    common_params_speculative_ngram_map ngram_map_k4v;
+
+    common_params_speculative_ngram_cache ngram_cache;
 
     bool has_dft() const {
-        return !mparams_dft.path.empty() || !mparams_dft.hf_repo.empty();
+        return !draft.mparams.path.empty() || !draft.mparams.hf_repo.empty();
     }
 };
 
 struct common_params_vocoder {
     struct common_params_model model;
 
-    std::string speaker_file = ""; // speaker file path                                      // NOLINT
+    std::string speaker_file; // speaker file path
 
-    bool use_guide_tokens = false; // enable guide tokens to improve TTS accuracy            // NOLINT
+    bool use_guide_tokens = false; // enable guide tokens to improve TTS accuracy
 };
 
 struct common_params_diffusion {
@@ -433,19 +441,20 @@ struct common_params {
     // offload params
     std::vector<ggml_backend_dev_t> devices; // devices to use for offloading
 
-    int32_t n_gpu_layers       = -1;   // number of layers to store in VRAM, -1 is auto, <= -2 is all
-    int32_t main_gpu           = 0;    // the GPU that is used for scratch and small tensors
-    float   tensor_split[128]  = {0};  // how split tensors should be distributed across GPUs
-    bool    fit_params         = true; // whether to fit unset model/context parameters to free device memory
-    int32_t fit_params_min_ctx = 4096; // minimum context size to set when trying to reduce memory use
+    int32_t n_gpu_layers       = -1;    // number of layers to store in VRAM, -1 is auto, <= -2 is all
+    int32_t main_gpu           = 0;     // the GPU that is used for scratch and small tensors
+    float   tensor_split[128]  = {0};   // how split tensors should be distributed across GPUs
+    bool    fit_params         = true;  // whether to fit unset model/context parameters to free device memory
+    bool    fit_params_print   = false; // print the estimated required memory to run the model
+    int32_t fit_params_min_ctx = 4096;  // minimum context size to set when trying to reduce memory use
 
     // margin per device in bytes for fitting parameters to free memory:
     std::vector<size_t> fit_params_target = std::vector<size_t>(llama_max_devices(), 1024 * 1024*1024);
 
     enum llama_split_mode split_mode = LLAMA_SPLIT_MODE_LAYER; // how to split the model across GPUs
 
-    struct cpu_params cpuparams;
-    struct cpu_params cpuparams_batch;
+    common_cpu_params cpuparams;
+    common_cpu_params cpuparams_batch;
 
     ggml_backend_sched_eval_callback cb_eval = nullptr;
     void * cb_eval_user_data                 = nullptr;
@@ -579,7 +588,7 @@ struct common_params {
     int32_t n_threads_http      = -1;    // number of threads to process HTTP requests (TODO: support threadpool)
     int32_t n_cache_reuse       = 0;     // min chunk size to reuse from the cache via KV shifting
     bool    cache_prompt        = true;  // whether to enable prompt caching
-    bool    clear_idle          = true;  // save and clear idle slots upon starting a new task
+    bool    cache_idle_slots    = true;  // save and clear idle slots upon starting a new task
     int32_t n_ctx_checkpoints   = 32;    // max number of context checkpoints per slot
     int32_t checkpoint_every_nt = 8192;  // make a checkpoint every n tokens during prefill
     int32_t cache_ram_mib       = 8192;  // -1 = no limit, 0 - disable, 1 = 1 MiB, etc.
@@ -593,8 +602,6 @@ struct common_params {
     bool force_pure_content_parser = false;
     common_reasoning_format reasoning_format = COMMON_REASONING_FORMAT_DEEPSEEK;
     int enable_reasoning = -1; // -1 = auto, 0 = disable, 1 = enable
-    int reasoning_budget = -1;
-    std::string reasoning_budget_message; // message injected before end tag when budget exhausted
     bool prefill_assistant = true; // if true, any trailing assistant message will be prefilled into the response
     int sleep_idle_seconds = -1;   // if >0, server will sleep after this many seconds of idle time
 
@@ -691,7 +698,7 @@ std::string common_params_get_system_info(const common_params & params);
 
 bool parse_cpu_range(const std::string & range, bool(&boolmask)[GGML_MAX_N_THREADS]);
 bool parse_cpu_mask(const std::string & mask, bool(&boolmask)[GGML_MAX_N_THREADS]);
-void postprocess_cpu_params(cpu_params & cpuparams, const cpu_params * role_model = nullptr);
+void postprocess_cpu_params(common_cpu_params & cpuparams, const common_cpu_params * role_model = nullptr);
 bool set_process_priority(enum ggml_sched_priority prio);
 
 //
@@ -759,6 +766,11 @@ inline bool string_starts_with(std::string_view str, std::string_view prefix) {
            str.compare(0, prefix.size(), prefix) == 0;
 }
 
+// remove when moving to c++20
+inline bool string_starts_with(std::string_view str, char prefix) {
+    return !str.empty() && str.front() == prefix;
+}
+
 // remove when moving to c++20
 inline bool string_ends_with(std::string_view str, std::string_view suffix) {
     return str.size() >= suffix.size() &&
@@ -854,12 +866,28 @@ common_init_result_ptr common_init_from_params(common_params & params);
 
 struct llama_model_params     common_model_params_to_llama  (      common_params & params);
 struct llama_context_params   common_context_params_to_llama(const common_params & params);
-struct ggml_threadpool_params ggml_threadpool_params_from_cpu_params(const cpu_params & params);
+struct ggml_threadpool_params ggml_threadpool_params_from_cpu_params(const common_cpu_params & params);
 
 // clear LoRA adapters from context, then apply new list of adapters
 void common_set_adapter_lora(struct llama_context * ctx, std::vector<common_adapter_lora_info> & lora);
 
-std::string                   get_model_endpoint();
+// model endpoint from env
+std::string common_get_model_endpoint();
+
+//
+// Context utils
+//
+
+enum common_context_seq_rm_type {
+    COMMON_CONTEXT_SEQ_RM_TYPE_NO   = 0, // seq_rm not supported (e.g. no memory module)
+    COMMON_CONTEXT_SEQ_RM_TYPE_PART = 1, // can seq_rm partial sequences
+    COMMON_CONTEXT_SEQ_RM_TYPE_FULL = 2, // can seq_rm full sequences only
+};
+
+// check if the llama_context can remove sequences
+// note: clears the memory of the context
+common_context_seq_rm_type common_context_can_seq_rm(llama_context * ctx);
+
 
 //
 // Batch utils

common/debug.cpp

@@ -1,9 +1,38 @@
 #include "debug.h"
 
+#include "common.h"
 #include "log.h"
 
 #include <cmath>
+#include <regex>
 #include <string>
+#include <vector>
+
+struct common_debug_cb_user_data::impl {
+    std::vector<uint8_t>    data;
+    std::vector<std::regex> tensor_filters;
+    bool                    abort_on_nan{false};
+};
+
+common_debug_cb_user_data::common_debug_cb_user_data() : pimpl(std::make_unique<impl>()) {}
+common_debug_cb_user_data::~common_debug_cb_user_data() = default;
+
+common_debug_cb_user_data::common_debug_cb_user_data(common_params & params, const std::vector<std::string> & filter_patterns, bool abort_on_nan)
+    : pimpl(std::make_unique<impl>())
+{
+    for (const auto & pattern : filter_patterns) {
+        try {
+            std::string anchored_pattern = "^" + pattern;
+            pimpl->tensor_filters.emplace_back(anchored_pattern, std::regex::optimize);
+        } catch (const std::regex_error & e) {
+            throw std::runtime_error("Invalid regex pattern '" + pattern + "': " + e.what());
+        }
+    }
+    pimpl->abort_on_nan = abort_on_nan;
+
+    params.cb_eval           = common_debug_cb_eval;
+    params.cb_eval_user_data = this;
+}
 
 static std::string common_ggml_ne_string(const ggml_tensor * t) {
     std::string str;
@@ -47,8 +76,7 @@ static float common_ggml_get_float_value(const uint8_t * data,
 
 #define INDENT "    "
 
-template <bool abort>
-void common_debug_print_tensor(uint8_t * data, ggml_type type, const int64_t * ne, const size_t * nb, int64_t n) {
+static void common_debug_print_tensor(uint8_t * data, ggml_type type, const int64_t * ne, const size_t * nb, int64_t n, bool abort_on_nan) {
     GGML_ASSERT(n > 0);
     float sum = 0;
     for (int64_t i3 = 0; i3 < ne[3]; i3++) {
@@ -94,7 +122,7 @@ void common_debug_print_tensor(uint8_t * data, ggml_type type, const int64_t * n
         LOG(INDENT "sum = %f\n", sum);
     }
 
-    if constexpr (abort) {
+    if (abort_on_nan) {
         if (std::isnan(sum)) {
             LOG("encountered NaN - aborting\n");
             exit(0);
@@ -112,8 +140,9 @@ void common_debug_print_tensor(uint8_t * data, ggml_type type, const int64_t * n
  * @param user_data user data to pass at each call back
  * @return true to receive data or continue the graph, false otherwise
  */
-template <bool abort_on_nan> bool common_debug_cb_eval(struct ggml_tensor * t, bool ask, void * user_data) {
-    auto * cb_data = (base_callback_data *) user_data;
+bool common_debug_cb_eval(struct ggml_tensor * t, bool ask, void * user_data) {
+    auto * cb_data = (common_debug_cb_user_data *) user_data;
+    auto * pimpl = cb_data->pimpl.get();
 
     const struct ggml_tensor * src0 = t->src[0];
     const struct ggml_tensor * src1 = t->src[1];
@@ -122,10 +151,10 @@ template <bool abort_on_nan> bool common_debug_cb_eval(struct ggml_tensor * t, b
         return true;  // Always retrieve data
     }
 
-    bool matches_filter = cb_data->tensor_filters.empty();
+    bool matches_filter = pimpl->tensor_filters.empty();
 
     if (!matches_filter) {
-        for (const auto & filter : cb_data->tensor_filters) {
+        for (const auto & filter : pimpl->tensor_filters) {
             if (std::regex_search(t->name, filter)) {
                 matches_filter = true;
                 break;
@@ -148,20 +177,14 @@ template <bool abort_on_nan> bool common_debug_cb_eval(struct ggml_tensor * t, b
 
     if (!is_host) {
         auto n_bytes = ggml_nbytes(t);
-        cb_data->data.resize(n_bytes);
-        ggml_backend_tensor_get(t, cb_data->data.data(), 0, n_bytes);
+        pimpl->data.resize(n_bytes);
+        ggml_backend_tensor_get(t, pimpl->data.data(), 0, n_bytes);
     }
 
     if (!ggml_is_quantized(t->type) && matches_filter) {
-        uint8_t * data = is_host ? (uint8_t *) t->data : cb_data->data.data();
-        common_debug_print_tensor<abort_on_nan>(data, t->type, t->ne, t->nb, 3);
+        uint8_t * data = is_host ? (uint8_t *) t->data : pimpl->data.data();
+        common_debug_print_tensor(data, t->type, t->ne, t->nb, 3, pimpl->abort_on_nan);
     }
 
     return true;
 }
-
-// Explicit template instantiations
-template bool common_debug_cb_eval<false>(ggml_tensor *, bool, void *);
-template bool common_debug_cb_eval<true>(ggml_tensor *, bool, void *);
-template void common_debug_print_tensor<false>(uint8_t *, ggml_type, const int64_t *, const size_t *, int64_t);
-template void common_debug_print_tensor<true>(uint8_t *, ggml_type, const int64_t *, const size_t *, int64_t);

common/debug.h

@@ -1,43 +1,31 @@
 #pragma once
-#include "common.h"
+
+#include <memory>
 #include <string>
 #include <vector>
-#include <regex>
 
 // common debug functions and structs
 
-// Print a tensor's detailed data
-// data - the tensor's data in byte format
-// type - the tensor's quantization type
-// ne   - the tensor dimensions array
-// nb   - the tensor strides array
-// n    - the number of rows/columns to fully print
-template <bool abort_on_nan> void common_debug_print_tensor(uint8_t * data, ggml_type type, const int64_t * ne, const size_t * nb, int64_t n);
+struct common_params;
 
 // Intended to use as callback for ggml_backend_sched_eval_callback
 // prints tensors that are processed in the computation graph
-// by default prints all tensors, but can be configured by creating a `base_callback_data` instance with
-// non-empty filter_patterns. See examples/debug.ccp for possible usage patterns
-// The template parameter determines whether an error should be thrown whenever a NaN is encountered
+// by default prints all tensors, but can be configured by creating a `common_debug_cb_user_data` instance with
+// non-empty filter_patterns. See examples/debug.cpp for possible usage patterns
+// `common_debug_cb_user_data` contains `abort_on_nan` flag that determines whether an error should be thrown whenever a NaN is encountered
 // in a tensor (useful for stopping debug sessions on first erroneous tensor)
 // The callback data will be passed as the third parameter (user_data)
-template <bool abort_on_nan> bool common_debug_cb_eval(struct ggml_tensor * t, bool ask, void * user_data);
-struct base_callback_data {
-    std::vector<uint8_t>    data;
-    std::vector<std::regex> tensor_filters;
+bool common_debug_cb_eval(struct ggml_tensor * t, bool ask, void * user_data);
 
-    base_callback_data() = default;
+struct common_debug_cb_user_data {
+    struct impl;
+    std::unique_ptr<impl> pimpl;
 
-    base_callback_data(common_params & params, const std::vector<std::string> & filter_patterns) {
-        for (const auto & pattern : filter_patterns) {
-            try {
-                std::string anchored_pattern = "^" + pattern;
-                tensor_filters.emplace_back(anchored_pattern, std::regex::optimize);
-            } catch (const std::regex_error & e) {
-                throw std::runtime_error("Invalid regex pattern '" + pattern + "': " + e.what());
-            }
-        }
-        params.cb_eval           = common_debug_cb_eval<false>;
-        params.cb_eval_user_data = this;
-    }
+    common_debug_cb_user_data();
+    ~common_debug_cb_user_data();
+
+    common_debug_cb_user_data(const common_debug_cb_user_data &) = delete;
+    common_debug_cb_user_data & operator=(const common_debug_cb_user_data &) = delete;
+
+    common_debug_cb_user_data(common_params & params, const std::vector<std::string> & filter_patterns, bool abort_on_nan = false);
 };

common/download.cpp

@@ -1,5 +1,6 @@
 #include "arg.h"
 
+#include "build-info.h"
 #include "common.h"
 #include "log.h"
 #include "download.h"
@@ -303,7 +304,7 @@ static int common_download_file_single_online(const std::string & url,
         headers.emplace(h.first, h.second);
     }
     if (headers.find("User-Agent") == headers.end()) {
-        headers.emplace("User-Agent", "llama-cpp/" + build_info);
+        headers.emplace("User-Agent", "llama-cpp/" + std::string(llama_build_info()));
     }
     if (!opts.bearer_token.empty()) {
         headers.emplace("Authorization", "Bearer " + opts.bearer_token);
@@ -441,7 +442,7 @@ std::pair<long, std::vector<char>> common_remote_get_content(const std::string
         headers.emplace(h.first, h.second);
     }
     if (headers.find("User-Agent") == headers.end()) {
-        headers.emplace("User-Agent", "llama-cpp/" + build_info);
+        headers.emplace("User-Agent", "llama-cpp/" + std::string(llama_build_info()));
     }
 
     if (params.timeout > 0) {
@@ -626,7 +627,7 @@ static hf_cache::hf_file find_best_model(const hf_cache::hf_files & files,
     if (!tag.empty()) {
         tags.push_back(tag);
     } else {
-        tags = {"Q4_K_M", "Q4_0"};
+        tags = {"Q4_K_M", "Q8_0"};
     }
 
     for (const auto & t : tags) {

common/fit.cpp

@@ -0,0 +1,951 @@
+#include "fit.h"
+
+#include "log.h"
+
+#include "../src/llama-ext.h"
+
+#include <array>
+#include <cassert>
+#include <stdexcept>
+#include <cinttypes>
+#include <set>
+#include <string>
+#include <vector>
+
+// this enum is only used in llama_params_fit_impl but needs to be defined outside of it to fix a Windows compilation issue
+// enum to identify part of a layer for distributing its tensors:
+enum common_layer_fraction_t {
+    LAYER_FRACTION_NONE = 0, // nothing
+    LAYER_FRACTION_ATTN = 1, // attention
+    LAYER_FRACTION_UP   = 2, // attention + up
+    LAYER_FRACTION_GATE = 3, // attention + up + gate
+    LAYER_FRACTION_MOE  = 4, // everything but sparse MoE weights
+};
+
+class common_params_fit_exception : public std::runtime_error {
+    using std::runtime_error::runtime_error;
+};
+
+static std::vector<llama_device_memory_data> common_get_device_memory_data(
+        const char * path_model,
+        const llama_model_params * mparams,
+        const llama_context_params * cparams,
+        std::vector<ggml_backend_dev_t> & devs,
+        uint32_t & hp_ngl,
+        uint32_t & hp_n_ctx_train,
+        uint32_t & hp_n_expert,
+        ggml_log_level log_level) {
+    struct user_data_t {
+        struct {
+            ggml_log_callback callback;
+            void * user_data;
+        } original_logger;
+        ggml_log_level min_level; // prints below this log level go to debug log
+    };
+    user_data_t ud;
+    llama_log_get(&ud.original_logger.callback, &ud.original_logger.user_data);
+    ud.min_level = log_level;
+
+    llama_log_set([](ggml_log_level level, const char * text, void * user_data) {
+        const user_data_t * ud = (const user_data_t *) user_data;
+        const ggml_log_level level_eff = level >= ud->min_level ? level : GGML_LOG_LEVEL_DEBUG;
+        ud->original_logger.callback(level_eff, text, ud->original_logger.user_data);
+    }, &ud);
+
+    llama_model_params mparams_copy = *mparams;
+    mparams_copy.no_alloc  = true;
+    mparams_copy.use_mmap  = false;
+    mparams_copy.use_mlock = false;
+
+    llama_model * model = llama_model_load_from_file(path_model, mparams_copy);
+    if (model == nullptr) {
+        llama_log_set(ud.original_logger.callback, ud.original_logger.user_data);
+        throw std::runtime_error("failed to load model");
+    }
+
+    llama_context * ctx = llama_init_from_model(model, *cparams);
+    if (ctx == nullptr) {
+        llama_model_free(model);
+        llama_log_set(ud.original_logger.callback, ud.original_logger.user_data);
+        throw std::runtime_error("failed to create llama_context from model");
+    }
+
+    const size_t nd = llama_model_n_devices(model);
+    std::vector<llama_device_memory_data> ret(nd + 1);
+
+    llama_memory_breakdown memory_breakdown = llama_get_memory_breakdown(ctx);
+
+    for (const auto & [buft, mb] : memory_breakdown) {
+        if (ggml_backend_buft_is_host(buft)) {
+            ret.back().mb.model   += mb.model;
+            ret.back().mb.context += mb.context;
+            ret.back().mb.compute += mb.compute;
+            continue;
+        }
+
+        ggml_backend_dev_t dev = ggml_backend_buft_get_device(buft);
+        if (!dev) {
+            continue;
+        }
+        for (size_t i = 0; i < nd; i++) {
+            if (dev == llama_model_get_device(model, i)) {
+                ret[i].mb.model   += mb.model;
+                ret[i].mb.context += mb.context;
+                ret[i].mb.compute += mb.compute;
+                break;
+            }
+        }
+    }
+
+    {
+        ggml_backend_dev_t cpu_dev = ggml_backend_dev_by_type(GGML_BACKEND_DEVICE_TYPE_CPU);
+        if (cpu_dev == nullptr) {
+            throw std::runtime_error("no CPU backend found");
+        }
+        size_t free;
+        size_t total;
+        ggml_backend_dev_memory(cpu_dev, &free, &total);
+        ret.back().free  = free;
+        ret.back().total = total;
+    }
+    for (size_t i = 0; i < nd; i++) {
+        size_t free;
+        size_t total;
+        ggml_backend_dev_memory(llama_model_get_device(model, i), &free, &total);
+
+        // devices can return 0 bytes for free and total memory if they do not
+        // have any to report. in this case, we will use the host memory as a fallback
+        // fixes: https://github.com/ggml-org/llama.cpp/issues/18577
+        if (free == 0 && total == 0) {
+            free  = ret.back().free;
+            total = ret.back().total;
+        }
+        ret[i].free  = free;
+        ret[i].total = total;
+    }
+
+    devs.clear();
+    for (int i = 0; i < llama_model_n_devices(model); i++) {
+        devs.push_back(llama_model_get_device(model, i));
+    }
+
+    hp_ngl         = llama_model_n_layer(model);
+    hp_n_ctx_train = llama_model_n_ctx_train(model);
+    hp_n_expert    = llama_model_n_expert(model);
+
+    common_memory_breakdown_print(ctx);
+
+    llama_free(ctx);
+    llama_model_free(model);
+    llama_log_set(ud.original_logger.callback, ud.original_logger.user_data);
+
+    return ret;
+}
+
+static void common_params_fit_impl(
+        const char * path_model, struct llama_model_params * mparams, struct llama_context_params * cparams,
+        float * tensor_split, struct llama_model_tensor_buft_override * tensor_buft_overrides,
+        size_t * margins_s, uint32_t n_ctx_min, enum ggml_log_level log_level) {
+    if (mparams->split_mode == LLAMA_SPLIT_MODE_TENSOR) {
+        throw common_params_fit_exception("llama_params_fit is not implemented for SPLIT_MODE_TENSOR, abort");
+    }
+    constexpr int64_t MiB = 1024*1024;
+    typedef std::vector<llama_device_memory_data> dmds_t;
+    const llama_model_params default_mparams = llama_model_default_params();
+
+    std::vector<ggml_backend_dev_t> devs;
+    uint32_t hp_ngl = 0; // hparams.n_gpu_layers
+    uint32_t hp_nct = 0; // hparams.n_ctx_train
+    uint32_t hp_nex = 0; // hparams.n_expert
+
+    // step 1: get data for default parameters and check whether any changes are necessary in the first place
+
+    LOG_INF("%s: getting device memory data for initial parameters:\n", __func__);
+    const dmds_t dmds_full = common_get_device_memory_data(path_model, mparams, cparams, devs, hp_ngl, hp_nct, hp_nex, log_level);
+    const size_t nd = devs.size(); // number of devices
+
+    std::vector<int64_t> margins; // this function uses int64_t rather than size_t for memory sizes to more conveniently handle deficits
+    margins.reserve(nd);
+    if (nd == 0) {
+        margins.push_back(margins_s[0]);
+    } else {
+        for (size_t id = 0; id < nd; id++) {
+            margins.push_back(margins_s[id]);
+        }
+    }
+
+    std::vector<std::string> dev_names;
+    {
+        dev_names.reserve(nd);
+        size_t max_length = 0;
+        for (const auto & dev : devs) {
+            std::string name = ggml_backend_dev_name(dev);
+            name += " (";
+            name += ggml_backend_dev_description(dev);
+            name += ")";
+            dev_names.push_back(name);
+            max_length = std::max(max_length, name.length());
+        }
+        for (std::string & dn : dev_names) {
+            dn.insert(dn.end(), max_length - dn.length(), ' ');
+        }
+    }
+
+    int64_t sum_free            = 0;
+    int64_t sum_projected_free  = 0;
+    int64_t sum_projected_used  = 0;
+    int64_t sum_projected_model = 0;
+    std::vector<int64_t> projected_free_per_device;
+    projected_free_per_device.reserve(nd);
+
+    if (nd == 0) {
+        sum_projected_used = dmds_full.back().mb.total();
+        sum_free           = dmds_full.back().total;
+        sum_projected_free = sum_free - sum_projected_used;
+        LOG_INF("%s: projected to use %" PRId64 " MiB of host memory vs. %" PRId64 " MiB of total host memory\n",
+            __func__, sum_projected_used/MiB, sum_free/MiB);
+        if (sum_projected_free >= margins[0]) {
+            LOG_INF("%s: will leave %" PRId64 " >= %" PRId64 " MiB of system memory, no changes needed\n",
+                __func__, sum_projected_free/MiB, margins[0]/MiB);
+            return;
+        }
+    } else {
+        if (nd > 1) {
+            LOG_INF("%s: projected memory use with initial parameters [MiB]:\n", __func__);
+        }
+        for (size_t id = 0; id < nd; id++) {
+            const llama_device_memory_data & dmd = dmds_full[id];
+
+            const int64_t projected_used = dmd.mb.total();
+            const int64_t projected_free = dmd.free - projected_used;
+            projected_free_per_device.push_back(projected_free);
+
+            sum_free            += dmd.free;
+            sum_projected_used  += projected_used;
+            sum_projected_free  += projected_free;
+            sum_projected_model += dmd.mb.model;
+
+            if (nd > 1) {
+                LOG_INF("%s:   - %s: %6" PRId64 " total, %6" PRId64 " used, %6" PRId64 " free vs. target of %6" PRId64 "\n",
+                    __func__, dev_names[id].c_str(), dmd.total/MiB, projected_used/MiB, projected_free/MiB, margins[id]/MiB);
+            }
+        }
+        assert(sum_free >= 0 && sum_projected_used >= 0);
+        LOG_INF("%s: projected to use %" PRId64 " MiB of device memory vs. %" PRId64 " MiB of free device memory\n",
+            __func__, sum_projected_used/MiB, sum_free/MiB);
+        if (nd == 1) {
+            if (projected_free_per_device[0] >= margins[0]) {
+                LOG_INF("%s: will leave %" PRId64 " >= %" PRId64 " MiB of free device memory, no changes needed\n",
+                    __func__, projected_free_per_device[0]/MiB, margins[0]/MiB);
+                return;
+            }
+        } else {
+            bool changes_needed = false;
+            for (size_t id = 0; id < nd; id++) {
+                if (projected_free_per_device[id] < margins[id]) {
+                    changes_needed = true;
+                    break;
+                }
+            }
+            if (!changes_needed) {
+                LOG_INF("%s: targets for free memory can be met on all devices, no changes needed\n", __func__);
+                return;
+            }
+        }
+    }
+
+    // step 2: try reducing memory use by reducing the context size
+
+    {
+        int64_t global_surplus = sum_projected_free;
+        if (nd == 0) {
+            global_surplus -= margins[0];
+        } else {
+            for (size_t id = 0; id < nd; id++) {
+                global_surplus -= margins[id];
+            }
+        }
+        if (global_surplus < 0) {
+            if (nd <= 1) {
+                LOG_INF("%s: cannot meet free memory target of %" PRId64 " MiB, need to reduce device memory by %" PRId64 " MiB\n",
+                    __func__, margins[0]/MiB, -global_surplus/MiB);
+            } else {
+                LOG_INF(
+                    "%s: cannot meet free memory targets on all devices, need to use %" PRId64 " MiB less in total\n",
+                    __func__, -global_surplus/MiB);
+            }
+            if (cparams->n_ctx == 0) {
+                if (hp_nct > n_ctx_min) {
+                    int64_t sum_used_target = sum_free;
+                    if (nd == 0) {
+                        sum_used_target -= margins[0];
+                    } else {
+                        for (size_t id = 0; id < nd; id++) {
+                            sum_used_target -= margins[id];
+                        }
+                    }
+                    if (nd > 1) {
+                        // for multiple devices we need to be more conservative in terms of how much context we think can fit:
+                        //   - for dense models only whole layers can be assigned to devices
+                        //   - for MoE models only whole tensors can be assigned to devices, which we estimate to be <= 1/3 of a layer
+                        //   - on average we expect a waste of 0.5 layers/tensors per device
+                        //   - use slightly more than the expected average for nd devices to be safe
+                        const int64_t model_per_layer = sum_projected_model / std::min(uint32_t(mparams->n_gpu_layers), hp_ngl);
+                        sum_used_target -= (nd + 1) * model_per_layer / (hp_nex == 0 ? 2 : 6);
+                    }
+
+                    int64_t sum_projected_used_min_ctx = 0;
+                    cparams->n_ctx = n_ctx_min;
+                    const dmds_t dmds_min_ctx = common_get_device_memory_data(path_model, mparams, cparams, devs, hp_ngl, hp_nct, hp_nex, log_level);
+                    if (nd == 0) {
+                        sum_projected_used_min_ctx = dmds_min_ctx.back().mb.total();
+                    } else {
+                        for (size_t id = 0; id < nd; id++) {
+                            sum_projected_used_min_ctx += dmds_min_ctx[id].mb.total();
+                        }
+                    }
+                    if (sum_used_target > sum_projected_used_min_ctx) {
+                        // linear interpolation between minimum and maximum context size:
+                        cparams->n_ctx += (hp_nct - n_ctx_min) * (sum_used_target - sum_projected_used_min_ctx)
+                            / (sum_projected_used - sum_projected_used_min_ctx);
+                        cparams->n_ctx = std::max(cparams->n_ctx - cparams->n_ctx % 256, n_ctx_min); // round down context for CUDA backend
+
+                        const int64_t bytes_per_ctx = (sum_projected_used - sum_projected_used_min_ctx) / (hp_nct - n_ctx_min);
+                        const int64_t memory_reduction = (hp_nct - cparams->n_ctx) * bytes_per_ctx;
+                        LOG_INF("%s: context size reduced from %" PRIu32 " to %" PRIu32 " -> need %" PRId64 " MiB less memory in total\n",
+                            __func__, hp_nct, cparams->n_ctx, memory_reduction/MiB);
+                        if (nd <= 1) {
+                            LOG_INF("%s: entire model can be fit by reducing context\n", __func__);
+                            return;
+                        }
+                        LOG_INF("%s: entire model should be fit across devices by reducing context\n", __func__);
+                    } else {
+                        const int64_t memory_reduction = sum_projected_used - sum_projected_used_min_ctx;
+                        LOG_INF("%s: context size reduced from %" PRIu32 " to %" PRIu32 " -> need %" PRId64 " MiB less memory in total\n",
+                            __func__, hp_nct, cparams->n_ctx, memory_reduction/MiB);
+                    }
+                } else {
+                    if (n_ctx_min == UINT32_MAX) {
+                        LOG_INF("%s: user has requested full context size of %" PRIu32 " -> no change\n", __func__, hp_nct);
+                    } else {
+                        LOG_INF("%s: default model context size is %" PRIu32 " which is <= the min. context size of %" PRIu32 " -> no change\n",
+                            __func__, hp_nct, n_ctx_min);
+                    }
+                }
+            } else {
+                LOG_INF("%s: context size set by user to %" PRIu32 " -> no change\n", __func__, cparams->n_ctx);
+            }
+        }
+    }
+    if (nd == 0) {
+        throw common_params_fit_exception("was unable to fit model into system memory by reducing context, abort");
+    }
+
+    if (mparams->n_gpu_layers != default_mparams.n_gpu_layers) {
+        throw common_params_fit_exception("n_gpu_layers already set by user to " + std::to_string(mparams->n_gpu_layers) + ", abort");
+    }
+    if (nd > 1) {
+        if (!tensor_split) {
+            throw common_params_fit_exception("did not provide a buffer to write the tensor_split to, abort");
+        }
+        if (mparams->tensor_split) {
+            for (size_t id = 0; id < nd; id++) {
+                if (mparams->tensor_split[id] != 0.0f) {
+                    throw common_params_fit_exception("model_params::tensor_split already set by user, abort");
+                }
+            }
+        }
+        if (mparams->split_mode == LLAMA_SPLIT_MODE_ROW) {
+            throw common_params_fit_exception("changing weight allocation for LLAMA_SPLIT_MODE_ROW not implemented, abort");
+        }
+    }
+    if (!tensor_buft_overrides) {
+        throw common_params_fit_exception("did not provide buffer to set tensor_buft_overrides, abort");
+    }
+    if (mparams->tensor_buft_overrides && (mparams->tensor_buft_overrides->pattern || mparams->tensor_buft_overrides->buft)) {
+        throw common_params_fit_exception("model_params::tensor_buft_overrides already set by user, abort");
+    }
+
+    // step 3: iteratively fill the back to front with "dense" layers
+    //   - for a dense model simply fill full layers, giving each device a contiguous slice of the model
+    //   - for a MoE model, same as dense model but with all MoE tensors in system memory
+
+    // utility function that returns a static C string matching the tensors for a specific layer index and layer fraction:
+    auto get_overflow_pattern = [&](const size_t il, const common_layer_fraction_t lf) -> const char * {
+        constexpr size_t n_strings = 1000;
+        if (il >= n_strings) {
+            throw std::runtime_error("at most " + std::to_string(n_strings) + " model layers are supported");
+        }
+        switch (lf) {
+            case LAYER_FRACTION_ATTN: {
+                static std::array<std::string, n_strings> patterns;
+                if (patterns[il].empty()) {
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(gate|up|gate_up|down).*";
+                }
+                return patterns[il].c_str();
+            }
+            case LAYER_FRACTION_UP: {
+                static std::array<std::string, n_strings> patterns;
+                if (patterns[il].empty()) {
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(gate|gate_up|down).*";
+                }
+                return patterns[il].c_str();
+            }
+            case LAYER_FRACTION_GATE: {
+                static std::array<std::string, n_strings> patterns;
+                if (patterns[il].empty()) {
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_down.*";
+                }
+                return patterns[il].c_str();
+            }
+            case LAYER_FRACTION_MOE: {
+                static std::array<std::string, n_strings> patterns;
+                if (patterns[il].empty()) {
+                    patterns[il] = "blk\\." + std::to_string(il) + "\\.ffn_(up|down|gate_up|gate)_(ch|)exps";
+                }
+                return patterns[il].c_str();
+            }
+            default:
+                GGML_ABORT("fatal error");
+        }
+    };
+
+    struct ngl_t {
+        uint32_t n_layer = 0; // number of total layers
+        uint32_t n_part  = 0; // number of partial layers, <= n_layer
+
+        // for the first partial layer varying parts can overflow, all further layers use LAYER_FRACTION_MOE:
+        common_layer_fraction_t overflow_type = LAYER_FRACTION_MOE;
+
+        uint32_t n_full() const {
+            assert(n_layer >= n_part);
+            return n_layer - n_part;
+        }
+    };
+
+    const size_t ntbo = llama_max_tensor_buft_overrides();
+
+    // utility function to set n_gpu_layers and tensor_split
+    auto set_ngl_tensor_split_tbo = [&](
+            const std::vector<ngl_t> & ngl_per_device,
+            const std::vector<ggml_backend_buffer_type_t> & overflow_bufts,
+            llama_model_params & mparams) {
+        mparams.n_gpu_layers = 0;
+        for (size_t id = 0; id < nd; id++) {
+            mparams.n_gpu_layers += ngl_per_device[id].n_layer;
+            if (nd > 1) {
+                tensor_split[id] = ngl_per_device[id].n_layer;
+            }
+        }
+        assert(uint32_t(mparams.n_gpu_layers) <= hp_ngl + 1);
+        uint32_t il0 = hp_ngl + 1 - mparams.n_gpu_layers; // start index for tensor buft overrides
+
+        mparams.tensor_split = tensor_split;
+
+        size_t itbo = 0;
+        for (size_t id = 0; id < nd; id++) {
+            il0 += ngl_per_device[id].n_full();
+            for (uint32_t il = il0; il < il0 + ngl_per_device[id].n_part; il++) {
+                if (itbo + 1 >= ntbo) {
+                    tensor_buft_overrides[itbo].pattern = nullptr;
+                    tensor_buft_overrides[itbo].buft    = nullptr;
+                    itbo++;
+                    mparams.tensor_buft_overrides = tensor_buft_overrides;
+                    throw common_params_fit_exception("llama_max_tensor_buft_overrides() == "
+                        + std::to_string(ntbo) + " is insufficient for model");
+                }
+                tensor_buft_overrides[itbo].pattern = get_overflow_pattern(il, il == il0 ? ngl_per_device[id].overflow_type : LAYER_FRACTION_MOE);
+                tensor_buft_overrides[itbo].buft = il == il0 ? overflow_bufts[id] : ggml_backend_cpu_buffer_type();
+                itbo++;
+            }
+            il0 += ngl_per_device[id].n_part;
+        }
+        tensor_buft_overrides[itbo].pattern = nullptr;
+        tensor_buft_overrides[itbo].buft    = nullptr;
+        itbo++;
+        mparams.tensor_buft_overrides = tensor_buft_overrides;
+    };
+
+    // utility function that returns the memory use per device for given numbers of layers per device
+    auto get_memory_for_layers = [&](
+            const char * func_name,
+            const std::vector<ngl_t> & ngl_per_device,
+            const std::vector<ggml_backend_buffer_type_t> & overflow_bufts) -> std::vector<int64_t> {
+        llama_model_params mparams_copy = *mparams;
+        set_ngl_tensor_split_tbo(ngl_per_device, overflow_bufts, mparams_copy);
+
+        const dmds_t dmd_nl = common_get_device_memory_data(
+            path_model, &mparams_copy, cparams, devs, hp_ngl, hp_nct, hp_nex, log_level);
+
+        LOG_INF("%s: memory for test allocation by device:\n", func_name);
+        for (size_t id = 0; id < nd; id++) {
+            const ngl_t & n = ngl_per_device[id];
+            LOG_INF(
+                "%s: id=%zu, n_layer=%2" PRIu32 ", n_part=%2" PRIu32 ", overflow_type=%d, mem=%6" PRId64 " MiB\n",
+                func_name, id, n.n_layer, n.n_part, int(n.overflow_type), dmd_nl[id].mb.total()/MiB);
+        }
+
+        std::vector<int64_t> ret;
+        ret.reserve(nd);
+        for (size_t id = 0; id < nd; id++) {
+            ret.push_back(dmd_nl[id].mb.total());
+        }
+        return ret;
+    };
+
+    int64_t global_surplus_cpu_moe = 0;
+    if (hp_nex > 0) {
+        const static std::string pattern_moe_all = "blk\\.\\d+\\.ffn_(up|down|gate_up|gate)_(ch|)exps"; // matches all MoE tensors
+        ggml_backend_buffer_type_t cpu_buft = ggml_backend_cpu_buffer_type();
+        tensor_buft_overrides[0] = {pattern_moe_all.c_str(), cpu_buft};
+        tensor_buft_overrides[1] = {nullptr, nullptr};
+        mparams->tensor_buft_overrides = tensor_buft_overrides;
+
+        LOG_INF("%s: getting device memory data with all MoE tensors moved to system memory:\n", __func__);
+        const dmds_t dmds_cpu_moe = common_get_device_memory_data(
+            path_model, mparams, cparams, devs, hp_ngl, hp_nct, hp_nex, log_level);
+
+        for (size_t id = 0; id < nd; id++) {
+            global_surplus_cpu_moe += dmds_cpu_moe[id].free;
+            global_surplus_cpu_moe -= int64_t(dmds_cpu_moe[id].mb.total()) + margins[id];
+        }
+
+        if (global_surplus_cpu_moe > 0) {
+            LOG_INF("%s: with only dense weights in device memory there is a total surplus of %" PRId64 " MiB\n",
+                __func__, global_surplus_cpu_moe/MiB);
+        } else {
+            LOG_INF("%s: with only dense weights in device memory there is still a total deficit of %" PRId64 " MiB\n",
+                __func__, -global_surplus_cpu_moe/MiB);
+        }
+
+        // reset
+        tensor_buft_overrides[0] = {nullptr, nullptr};
+        mparams->tensor_buft_overrides = tensor_buft_overrides;
+    }
+
+    std::vector<int64_t> targets; // maximum acceptable memory use per device
+    targets.reserve(nd);
+    for (size_t id = 0; id < nd; id++) {
+        targets.push_back(dmds_full[id].free - margins[id]);
+        LOG_INF("%s: id=%zu, target=%" PRId64 " MiB\n", __func__, id, targets[id]/MiB);
+    }
+
+    std::vector<ggml_backend_buffer_type_t> overflow_bufts; // which bufts the first partial layer of a device overflows to:
+    overflow_bufts.reserve(nd);
+    for (size_t id = 0; id < nd; id++) {
+        overflow_bufts.push_back(ggml_backend_cpu_buffer_type());
+    }
+
+    std::vector<ngl_t> ngl_per_device(nd);
+    std::vector<int64_t> mem = get_memory_for_layers(__func__, ngl_per_device, overflow_bufts);
+
+    // optimize the number of layers per device using the method of false position:
+    //   - ngl_per_device has 0 layers for each device, lower bound
+    //   - try a "high" configuration where a device is given all unassigned layers
+    //   - interpolate the memory use / layer between low and high linearly to get a guess where it meets our target
+    //   - check memory use of our guess, replace either the low or high bound
+    //   - once we only have a difference of a single layer, stop and return the lower bound that just barely still fits
+    //   - the last device has the output layer, which cannot be a partial layer
+    if (hp_nex == 0) {
+        LOG_INF("%s: filling dense layers back-to-front:\n", __func__);
+    } else {
+        LOG_INF("%s: filling dense-only layers back-to-front:\n", __func__);
+    }
+    for (int id = nd - 1; id >= 0; id--) {
+        uint32_t n_unassigned = hp_ngl + 1;
+        for (size_t jd = id + 1; jd < nd; ++jd) {
+            assert(n_unassigned >= ngl_per_device[jd].n_layer);
+            n_unassigned -= ngl_per_device[jd].n_layer;
+        }
+
+        std::vector<ngl_t> ngl_per_device_high = ngl_per_device;
+        ngl_per_device_high[id].n_layer = n_unassigned;
+        if (hp_nex > 0) {
+            ngl_per_device_high[id].n_part = size_t(id) < nd - 1 ? ngl_per_device_high[id].n_layer : ngl_per_device_high[id].n_layer - 1;
+        }
+        if (ngl_per_device_high[id].n_layer > 0) {
+            std::vector<int64_t> mem_high = get_memory_for_layers(__func__, ngl_per_device_high, overflow_bufts);
+            if (mem_high[id] > targets[id]) {
+                assert(ngl_per_device_high[id].n_layer > ngl_per_device[id].n_layer);
+                uint32_t delta = ngl_per_device_high[id].n_layer - ngl_per_device[id].n_layer;
+                LOG_INF("%s: start filling device %" PRIu32 ", delta=%" PRIu32 "\n", __func__, id, delta);
+                while (delta > 1) {
+                    uint32_t step_size = int64_t(delta) * (targets[id] - mem[id]) / (mem_high[id] - mem[id]);
+                    step_size = std::max(step_size, uint32_t(1));
+                    step_size = std::min(step_size, delta - 1);
+
+                    std::vector<ngl_t> ngl_per_device_test = ngl_per_device;
+                    ngl_per_device_test[id].n_layer += step_size;
+                    if (hp_nex) {
+                        ngl_per_device_test[id].n_part += size_t(id) == nd - 1 && ngl_per_device_test[id].n_part == 0 ?
+                            step_size - 1 : step_size; // the first layer is the output layer which must always be full
+                    }
+                    const std::vector<int64_t> mem_test = get_memory_for_layers(__func__, ngl_per_device_test, overflow_bufts);
+
+                    if (mem_test[id] <= targets[id]) {
+                        ngl_per_device = ngl_per_device_test;
+                        mem            = mem_test;
+                        LOG_INF("%s: set ngl_per_device[%d].n_layer=%" PRIu32 "\n", __func__, id, ngl_per_device[id].n_layer);
+                    } else {
+                        ngl_per_device_high = ngl_per_device_test;
+                        mem_high            = mem_test;
+                        LOG_INF("%s: set ngl_per_device_high[%d].n_layer=%" PRIu32 "\n", __func__, id, ngl_per_device_high[id].n_layer);
+                    }
+                    delta = ngl_per_device_high[id].n_layer - ngl_per_device[id].n_layer;
+                }
+            } else {
+                assert(ngl_per_device_high[id].n_layer == n_unassigned);
+                ngl_per_device = ngl_per_device_high;
+                mem            = mem_high;
+                LOG_INF("%s: set ngl_per_device[%d].n_layer=%" PRIu32 "\n", __func__, id, ngl_per_device[id].n_layer);
+            }
+        }
+
+        const int64_t projected_margin = dmds_full[id].free - mem[id];
+        LOG_INF(
+            "%s:   - %s: %2" PRIu32 " layers, %6" PRId64 " MiB used, %6" PRId64 " MiB free\n",
+            __func__, dev_names[id].c_str(), ngl_per_device[id].n_layer, mem[id]/MiB, projected_margin/MiB);
+    }
+    if (hp_nex == 0 || global_surplus_cpu_moe <= 0) {
+        set_ngl_tensor_split_tbo(ngl_per_device, overflow_bufts, *mparams);
+        return;
+    }
+
+    // step 4: for a MoE model where all dense tensors fit,
+    //     convert the dense-only layers in the back to full layers in the front until all devices are full
+    // essentially the same procedure as for the dense-only layers except front-to-back
+    // also, try fitting at least part of one more layer to reduce waste for "small" GPUs with e.g. 24 GiB VRAM
+
+    size_t id_dense_start = nd;
+    for (int id = nd - 1; id >= 0; id--) {
+        if (ngl_per_device[id].n_layer > 0) {
+            id_dense_start = id;
+            continue;
+        }
+        break;
+    }
+    assert(id_dense_start < nd);
+
+    LOG_INF("%s: converting dense-only layers to full layers and filling them front-to-back with overflow to next device/system memory:\n", __func__);
+    for (size_t id = 0; id <= id_dense_start && id_dense_start < nd; id++) {
+        std::vector<ngl_t> ngl_per_device_high = ngl_per_device;
+        for (size_t jd = id_dense_start; jd < nd; jd++) {
+            const uint32_t n_layer_move = jd < nd - 1 ? ngl_per_device_high[jd].n_layer : ngl_per_device_high[jd].n_layer - 1;
+            ngl_per_device_high[id].n_layer += n_layer_move;
+            ngl_per_device_high[jd].n_layer -= n_layer_move;
+            ngl_per_device_high[jd].n_part = 0;
+        }
+        size_t id_dense_start_high = nd - 1;
+        std::vector<int64_t> mem_high = get_memory_for_layers(__func__, ngl_per_device_high, overflow_bufts);
+
+        if (mem_high[id] > targets[id]) {
+            assert(ngl_per_device_high[id].n_full() >= ngl_per_device[id].n_full());
+            uint32_t delta = ngl_per_device_high[id].n_full() - ngl_per_device[id].n_full();
+            while (delta > 1) {
+                uint32_t step_size = int64_t(delta) * (targets[id] - mem[id]) / (mem_high[id] - mem[id]);
+                step_size = std::max(step_size, uint32_t(1));
+                step_size = std::min(step_size, delta - 1);
+
+                std::vector<ngl_t> ngl_per_device_test = ngl_per_device;
+                size_t id_dense_start_test = id_dense_start;
+                uint32_t n_converted_test = 0;
+                for (;id_dense_start_test < nd; id_dense_start_test++) {
+                    const uint32_t n_convert_jd = std::min(step_size - n_converted_test, ngl_per_device_test[id_dense_start_test].n_part);
+                    ngl_per_device_test[id_dense_start_test].n_layer -= n_convert_jd;
+                    ngl_per_device_test[id_dense_start_test].n_part -= n_convert_jd;
+                    ngl_per_device_test[id].n_layer += n_convert_jd;
+                    n_converted_test += n_convert_jd;
+
+                    if (ngl_per_device_test[id_dense_start_test].n_part > 0) {
+                        break;
+                    }
+                }
+                const std::vector<int64_t> mem_test = get_memory_for_layers(__func__, ngl_per_device_test, overflow_bufts);
+
+                if (mem_test[id] <= targets[id]) {
+                    ngl_per_device = ngl_per_device_test;
+                    mem            = mem_test;
+                    id_dense_start = id_dense_start_test;
+                    LOG_INF("%s: set ngl_per_device[%zu].(n_layer, n_part)=(%" PRIu32 ", %" PRIu32 "), id_dense_start=%zu\n",
+                        __func__, id, ngl_per_device[id].n_layer, ngl_per_device[id].n_part, id_dense_start);
+                } else {
+                    ngl_per_device_high = ngl_per_device_test;
+                    mem_high            = mem_test;
+                    id_dense_start_high = id_dense_start_test;
+                    LOG_INF("%s: set ngl_per_device_high[%zu].(n_layer, n_part)=(%" PRIu32 ", %" PRIu32 "), id_dense_start_high=%zu\n",
+                        __func__, id, ngl_per_device_high[id].n_layer, ngl_per_device_high[id].n_part, id_dense_start_high);
+                }
+                assert(ngl_per_device_high[id].n_full() >= ngl_per_device[id].n_full());
+                delta = ngl_per_device_high[id].n_full() - ngl_per_device[id].n_full();
+            }
+        } else {
+            ngl_per_device = ngl_per_device_high;
+            mem            = mem_high;
+            id_dense_start = id_dense_start_high;
+            LOG_INF("%s: set ngl_per_device[%zu].(n_layer, n_part)=(%" PRIu32 ", %" PRIu32 "), id_dense_start=%zu\n",
+                __func__, id, ngl_per_device[id].n_layer, ngl_per_device[id].n_part, id_dense_start);
+        }
+
+        // try to fit at least part of one more layer
+        if (ngl_per_device[id_dense_start].n_layer > (id < nd - 1 ? 0 : 1)) {
+            std::vector<ngl_t> ngl_per_device_test = ngl_per_device;
+            size_t id_dense_start_test = id_dense_start;
+            ngl_per_device_test[id_dense_start_test].n_layer--;
+            ngl_per_device_test[id_dense_start_test].n_part--;
+            ngl_per_device_test[id].n_layer++;
+            ngl_per_device_test[id].n_part++;
+            if (ngl_per_device_test[id_dense_start_test].n_part == 0) {
+                id_dense_start_test++;
+            }
+            ngl_per_device_test[id].overflow_type = LAYER_FRACTION_UP;
+            std::vector<ggml_backend_buffer_type_t> overflow_bufts_test = overflow_bufts;
+            if (id < nd - 1) {
+                overflow_bufts_test[id] = ggml_backend_dev_buffer_type(devs[id + 1]);
+            }
+            LOG_INF("%s: trying to fit one extra layer with overflow_type=LAYER_FRACTION_UP\n", __func__);
+            std::vector<int64_t> mem_test = get_memory_for_layers(__func__, ngl_per_device_test, overflow_bufts_test);
+            if (mem_test[id] < targets[id] && (id + 1 == nd || mem_test[id + 1] < targets[id + 1])) {
+                ngl_per_device = ngl_per_device_test;
+                overflow_bufts = overflow_bufts_test;
+                mem            = mem_test;
+                id_dense_start = id_dense_start_test;
+                LOG_INF("%s: set ngl_per_device[%zu].(n_layer, n_part, overflow_type)=(%" PRIu32 ", %" PRIu32 ", UP), id_dense_start=%zu\n",
+                    __func__, id, ngl_per_device[id].n_layer, ngl_per_device[id].n_part, id_dense_start);
+
+                ngl_per_device_test[id].overflow_type = LAYER_FRACTION_GATE;
+                LOG_INF("%s: trying to fit one extra layer with overflow_type=LAYER_FRACTION_GATE\n", __func__);
+                mem_test = get_memory_for_layers(__func__, ngl_per_device_test, overflow_bufts_test);
+                if (mem_test[id] < targets[id] && (id + 1 == nd || mem_test[id + 1] < targets[id + 1])) {
+                    ngl_per_device = ngl_per_device_test;
+                    overflow_bufts = overflow_bufts_test;
+                    mem            = mem_test;
+                    id_dense_start = id_dense_start_test;
+                    LOG_INF("%s: set ngl_per_device[%zu].(n_layer, n_part, overflow_type)=(%" PRIu32 ", %" PRIu32 ", GATE), id_dense_start=%zu\n",
+                        __func__, id, ngl_per_device[id].n_layer, ngl_per_device[id].n_part, id_dense_start);
+                }
+            } else {
+                ngl_per_device_test[id].overflow_type = LAYER_FRACTION_ATTN;
+                LOG_INF("%s: trying to fit one extra layer with overflow_type=LAYER_FRACTION_ATTN\n", __func__);
+                mem_test = get_memory_for_layers(__func__, ngl_per_device_test, overflow_bufts_test);
+                if (mem_test[id] < targets[id] && (id + 1 == nd || mem_test[id + 1] < targets[id + 1])) {
+                    ngl_per_device = ngl_per_device_test;
+                    overflow_bufts = overflow_bufts_test;
+                    mem            = mem_test;
+                    id_dense_start = id_dense_start_test;
+                    LOG_INF("%s: set ngl_per_device[%zu].(n_layer, n_part, overflow_type)=(%" PRIu32 ", %" PRIu32 ", ATTN), id_dense_start=%zu\n",
+                        __func__, id, ngl_per_device[id].n_layer, ngl_per_device[id].n_part, id_dense_start);
+                }
+            }
+        }
+
+        const int64_t projected_margin = dmds_full[id].free - mem[id];
+        LOG_INF(
+            "%s:   - %s: %2" PRIu32 " layers (%2" PRIu32 " overflowing), %6" PRId64 " MiB used, %6" PRId64 " MiB free\n",
+            __func__, dev_names[id].c_str(), ngl_per_device[id].n_layer, ngl_per_device[id].n_part, mem[id]/MiB, projected_margin/MiB);
+    }
+
+    // print info for devices that were not changed during the conversion from dense only to full layers:
+    for (size_t id = id_dense_start + 1; id < nd; id++) {
+        const int64_t projected_margin = dmds_full[id].free - mem[id];
+        LOG_INF(
+            "%s:   - %s: %2" PRIu32 " layers (%2" PRIu32 " overflowing), %6" PRId64 " MiB used, %6" PRId64 " MiB free\n",
+            __func__, dev_names[id].c_str(), ngl_per_device[id].n_layer, ngl_per_device[id].n_part, mem[id]/MiB, projected_margin/MiB);
+    }
+
+    set_ngl_tensor_split_tbo(ngl_per_device, overflow_bufts, *mparams);
+}
+
+enum common_params_fit_status common_fit_params(
+        const char * path_model,
+        llama_model_params * mparams,
+        llama_context_params * cparams,
+        float * tensor_split,
+        llama_model_tensor_buft_override * tensor_buft_overrides,
+        size_t * margins,
+        uint32_t n_ctx_min,
+        ggml_log_level log_level) {
+    const int64_t t0_us = llama_time_us();
+    common_params_fit_status status = COMMON_PARAMS_FIT_STATUS_SUCCESS;
+    try {
+        common_params_fit_impl(path_model, mparams, cparams, tensor_split, tensor_buft_overrides, margins, n_ctx_min, log_level);
+        LOG_INF("%s: successfully fit params to free device memory\n", __func__);
+    } catch (const common_params_fit_exception & e) {
+        LOG_WRN("%s: failed to fit params to free device memory: %s\n", __func__, e.what());
+        status = COMMON_PARAMS_FIT_STATUS_FAILURE;
+    } catch (const std::runtime_error & e) {
+        LOG_ERR("%s: encountered an error while trying to fit params to free device memory: %s\n", __func__, e.what());
+        status = COMMON_PARAMS_FIT_STATUS_ERROR;
+    }
+    const int64_t t1_us = llama_time_us();
+    LOG_INF("%s: fitting params to free memory took %.2f seconds\n", __func__, (t1_us - t0_us) * 1e-6);
+    return status;
+}
+
+void common_memory_breakdown_print(const struct llama_context * ctx) {
+    //const auto & devices = ctx->get_model().devices;
+    const auto * model = llama_get_model(ctx);
+
+    std::vector<ggml_backend_dev_t> devices;
+    for (int i = 0; i < llama_model_n_devices(model); i++) {
+        devices.push_back(llama_model_get_device(model, i));
+    }
+
+    llama_memory_breakdown memory_breakdown = llama_get_memory_breakdown(ctx);
+
+    std::vector<std::array<std::string, 9>> table_data;
+    table_data.reserve(devices.size());
+    const std::string template_header = "%s: | %s | %s   %s    %s   %s   %s   %s    %s |\n";
+    const std::string template_gpu    = "%s: | %s | %s = %s + (%s = %s + %s + %s) + %s |\n";
+    const std::string template_other  = "%s: | %s | %s   %s    %s = %s + %s + %s    %s |\n";
+
+    table_data.push_back({template_header, "memory breakdown [MiB]", "total", "free", "self", "model", "context", "compute", "unaccounted"});
+
+    constexpr size_t MiB = 1024 * 1024;
+    const std::vector<std::string> desc_prefixes_strip = {"NVIDIA ", "GeForce ", "Tesla ", "AMD ", "Radeon ", "Instinct "};
+
+    // track seen buffer types to avoid double counting:
+    std::set<ggml_backend_buffer_type_t> seen_buffer_types;
+
+    // accumulative memory breakdown for each device and for host:
+    std::vector<llama_memory_breakdown_data> mb_dev(devices.size());
+    llama_memory_breakdown_data              mb_host;
+
+    for (const auto & buft_mb : memory_breakdown) {
+        ggml_backend_buffer_type_t          buft = buft_mb.first;
+        const llama_memory_breakdown_data & mb   = buft_mb.second;
+        if (ggml_backend_buft_is_host(buft)) {
+            mb_host.model   += mb.model;
+            mb_host.context += mb.context;
+            mb_host.compute += mb.compute;
+            seen_buffer_types.insert(buft);
+            continue;
+        }
+        ggml_backend_dev_t dev = ggml_backend_buft_get_device(buft);
+        if (dev) {
+            int i_dev = -1;
+            for (size_t i = 0; i < devices.size(); i++) {
+                if (devices[i] == dev) {
+                    i_dev = i;
+                    break;
+                }
+            }
+            if (i_dev != -1) {
+                mb_dev[i_dev].model   += mb.model;
+                mb_dev[i_dev].context += mb.context;
+                mb_dev[i_dev].compute += mb.compute;
+                seen_buffer_types.insert(buft);
+                continue;
+            }
+        }
+    }
+
+    // print memory breakdown for each device:
+    for (size_t i = 0; i < devices.size(); i++) {
+        ggml_backend_dev_t dev = devices[i];
+        llama_memory_breakdown_data mb = mb_dev[i];
+
+        const std::string name = ggml_backend_dev_name(dev);
+        std::string desc = ggml_backend_dev_description(dev);
+        for (const std::string & prefix : desc_prefixes_strip) {
+            if (desc.length() >= prefix.length() && desc.substr(0, prefix.length()) == prefix) {
+                desc = desc.substr(prefix.length());
+            }
+        }
+
+        size_t free, total;
+        ggml_backend_dev_memory(dev, &free, &total);
+
+        const size_t self = mb.model + mb.context + mb.compute;
+        const int64_t unaccounted = static_cast<int64_t>(total) - static_cast<int64_t>(free) - static_cast<int64_t>(self);
+
+        table_data.push_back({
+            template_gpu,
+            "  - " + name + " (" + desc + ")",
+            std::to_string(total / MiB),
+            std::to_string(free / MiB),
+            std::to_string(self / MiB),
+            std::to_string(mb.model / MiB),
+            std::to_string(mb.context / MiB),
+            std::to_string(mb.compute / MiB),
+            std::to_string(unaccounted / static_cast<int64_t>(MiB))});
+    }
+
+    // print memory breakdown for host:
+    {
+        const size_t self = mb_host.model + mb_host.context + mb_host.compute;
+        table_data.push_back({
+            template_other,
+            "  - Host",
+            "", // total
+            "", // free
+            std::to_string(self / MiB),
+            std::to_string(mb_host.model / MiB),
+            std::to_string(mb_host.context / MiB),
+            std::to_string(mb_host.compute / MiB),
+            ""}); // unaccounted
+    }
+
+    // print memory breakdown for all remaining buffer types:
+    for (const auto & buft_mb : memory_breakdown) {
+        ggml_backend_buffer_type_t          buft = buft_mb.first;
+        const llama_memory_breakdown_data & mb   = buft_mb.second;
+        if (seen_buffer_types.count(buft) == 1) {
+            continue;
+        }
+        const std::string name = ggml_backend_buft_name(buft);
+        const size_t self = mb.model + mb.context + mb.compute;
+        table_data.push_back({
+            template_other,
+            "  - " + name,
+            "", // total
+            "", // free
+            std::to_string(self / MiB),
+            std::to_string(mb.model / MiB),
+            std::to_string(mb.context / MiB),
+            std::to_string(mb.compute / MiB),
+            ""}); // unaccounted
+        seen_buffer_types.insert(buft);
+    }
+
+    for (size_t j = 1; j < table_data[0].size(); j++) {
+        size_t max_len = 0;
+        for (const auto & td : table_data) {
+            max_len = std::max(max_len, td[j].length());
+        }
+        for (auto & td : table_data) {
+            td[j].insert(j == 1 ? td[j].length() : 0, max_len - td[j].length(), ' ');
+        }
+    }
+    for (const auto & td : table_data) {
+        LOG_INF(td[0].c_str(),
+            __func__, td[1].c_str(), td[2].c_str(), td[3].c_str(), td[4].c_str(), td[5].c_str(),
+            td[6].c_str(), td[7].c_str(), td[8].c_str());
+    }
+}
+
+void common_fit_print(
+        const char * path_model,
+        llama_model_params * mparams,
+        llama_context_params * cparams) {
+    std::vector<ggml_backend_dev_t> devs;
+    uint32_t hp_ngl = 0; // hparams.n_gpu_layers
+    uint32_t hp_nct = 0; // hparams.n_ctx_train
+    uint32_t hp_nex = 0; // hparams.n_expert
+
+    auto dmd = common_get_device_memory_data(path_model, mparams, cparams, devs, hp_ngl, hp_nct, hp_nex, GGML_LOG_LEVEL_ERROR);
+    GGML_ASSERT(dmd.size() == devs.size() + 1);
+
+    for (size_t id = 0; id < devs.size(); id++) {
+        printf("%s ",  ggml_backend_dev_name(devs[id]));
+        printf("%zu ", dmd[id].mb.model/1024/1024);
+        printf("%zu ", dmd[id].mb.context/1024/1024);
+        printf("%zu ", dmd[id].mb.compute/1024/1024);
+        printf("\n");
+    }
+
+    printf("Host ");
+    printf("%zu ", dmd.back().mb.model/1024/1024);
+    printf("%zu ", dmd.back().mb.context/1024/1024);
+    printf("%zu ", dmd.back().mb.compute/1024/1024);
+    printf("\n");
+}

common/fit.h

@@ -0,0 +1,32 @@
+#pragma once
+
+#include "ggml.h"
+
+enum common_params_fit_status {
+    COMMON_PARAMS_FIT_STATUS_SUCCESS = 0, // found allocations that are projected to fit
+    COMMON_PARAMS_FIT_STATUS_FAILURE = 1, // could not find allocations that are projected to fit
+    COMMON_PARAMS_FIT_STATUS_ERROR   = 2, // a hard error occurred, e.g. because no model could be found at the specified path
+};
+
+// fits mparams and cparams to free device memory (assumes system memory is unlimited)
+//   - returns true if the parameters could be successfully modified to fit device memory
+//   - this function is NOT thread safe because it modifies the global llama logger state
+//   - only parameters that have the same value as in llama_default_model_params are modified
+//     with the exception of the context size which is modified if and only if equal to 0
+enum common_params_fit_status common_fit_params(
+                               const char   * path_model,
+                struct llama_model_params   * mparams,
+                struct llama_context_params * cparams,
+                                      float * tensor_split,          // writable buffer for tensor split, needs at least llama_max_devices elements
+    struct llama_model_tensor_buft_override * tensor_buft_overrides, // writable buffer for overrides, needs at least llama_max_tensor_buft_overrides elements
+                                     size_t * margins,               // margins of memory to leave per device in bytes
+                                   uint32_t   n_ctx_min,             // minimum context size to set when trying to reduce memory use
+                        enum ggml_log_level   log_level);            // minimum log level to print during fitting, lower levels go to debug log
+
+// print estimated memory to stdout
+void common_fit_print(
+                               const char   * path_model,
+                struct llama_model_params   * mparams,
+                struct llama_context_params * cparams);
+
+void common_memory_breakdown_print(const struct llama_context * ctx);

common/hf-cache.cpp

@@ -1,5 +1,6 @@
 #include "hf-cache.h"
 
+#include "build-info.h"
 #include "common.h"
 #include "log.h"
 #include "http.h"
@@ -56,7 +57,7 @@ static fs::path get_cache_directory() {
 #ifndef _WIN32
         const struct passwd * pw = getpwuid(getuid());
 
-        if (pw->pw_dir && *pw->pw_dir) {
+        if (pw && pw->pw_dir && *pw->pw_dir) {
             return fs::path(pw->pw_dir) / ".cache" / "huggingface" / "hub";
         }
 #endif
@@ -200,7 +201,7 @@ static nl::json api_get(const std::string & url,
     auto [cli, parts] = common_http_client(url);
 
     httplib::Headers headers = {
-        {"User-Agent", "llama-cpp/" + build_info},
+        {"User-Agent", "llama-cpp/" + std::string(llama_build_info())},
         {"Accept", "application/json"}
     };
 
@@ -229,7 +230,7 @@ static nl::json api_get(const std::string & url,
 static std::string get_repo_commit(const std::string & repo_id,
                                    const std::string & token) {
     try {
-        auto endpoint = get_model_endpoint();
+        auto endpoint = common_get_model_endpoint();
         auto json = api_get(endpoint + "api/models/" + repo_id + "/refs", token);
 
         if (!json.is_object() ||
@@ -307,7 +308,7 @@ hf_files get_repo_files(const std::string & repo_id,
     hf_files files;
 
     try {
-        auto endpoint = get_model_endpoint();
+        auto endpoint = common_get_model_endpoint();
         auto json = api_get(endpoint + "api/models/" + repo_id + "/tree/" + commit + "?recursive=true", token);
 
         if (!json.is_array()) {

common/jinja/caps.cpp

@@ -1,4 +1,3 @@
-#include "log.h"
 #include "value.h"
 #include "runtime.h"
 #include "caps.h"

common/jinja/runtime.h

@@ -106,10 +106,16 @@ struct statement {
     size_t pos; // position in source, for debugging
     virtual ~statement() = default;
     virtual std::string type() const { return "Statement"; }
+
     // execute_impl must be overridden by derived classes
-    virtual value execute_impl(context &) { throw std::runtime_error("cannot exec " + type()); }
+    virtual value execute_impl(context &) { throw_exec_error(); }
     // execute is the public method to execute a statement with error handling
     value execute(context &);
+
+private:
+    [[noreturn]] void throw_exec_error() const {
+        throw std::runtime_error("cannot exec " + type());
+    }
 };
 
 // Type Checking Utilities
@@ -143,7 +149,7 @@ struct program : public statement {
     program() = default;
     explicit program(statements && body) : body(std::move(body)) {}
     std::string type() const override { return "Program"; }
-    value execute_impl(context &) override {
+    [[noreturn]] value execute_impl(context &) override {
         throw std::runtime_error("Cannot execute program directly, use jinja::runtime instead");
     }
 };
@@ -195,7 +201,7 @@ struct break_statement : public statement {
         }
     };
 
-    value execute_impl(context &) override {
+    [[noreturn]] value execute_impl(context &) override {
         throw break_statement::signal();
     }
 };
@@ -209,7 +215,7 @@ struct continue_statement : public statement {
         }
     };
 
-    value execute_impl(context &) override {
+    [[noreturn]] value execute_impl(context &) override {
         throw continue_statement::signal();
     }
 };
@@ -509,7 +515,7 @@ struct slice_expression : public expression {
         chk_type<expression>(this->step_expr);
     }
     std::string type() const override { return "SliceExpression"; }
-    value execute_impl(context &) override {
+    [[noreturn]] value execute_impl(context &) override {
         throw std::runtime_error("must be handled by MemberExpression");
     }
 };

common/jinja/value.cpp

@@ -590,6 +590,10 @@ static bool string_endswith(const std::string & str, const std::string & suffix)
     return str.compare(str.length() - suffix.length(), suffix.length(), suffix) == 0;
 }
 
+[[noreturn]] static value string_join_not_implemented(const func_args &) {
+    throw not_implemented_exception("String join builtin not implemented");
+}
+
 const func_builtins & value_string_t::get_builtins() const {
     static const func_builtins builtins = {
         {"default", default_value},
@@ -851,9 +855,7 @@ const func_builtins & value_string_t::get_builtins() const {
             res->val_str.mark_input_based_on(val_input->as_string());
             return res;
         }},
-        {"join", [](const func_args &) -> value {
-            throw not_implemented_exception("String join builtin not implemented");
-        }},
+        {"join", string_join_not_implemented},
     };
     return builtins;
 }
@@ -884,6 +886,9 @@ const func_builtins & value_bool_t::get_builtins() const {
     return builtins;
 }
 
+[[noreturn]] static value array_unique_not_implemented(const func_args &) {
+    throw not_implemented_exception("Array unique builtin not implemented");
+}
 
 const func_builtins & value_array_t::get_builtins() const {
     static const func_builtins builtins = {
@@ -1084,13 +1089,14 @@ const func_builtins & value_array_t::get_builtins() const {
             std::reverse(arr.begin(), arr.end());
             return is_val<value_tuple>(val) ? mk_val<value_tuple>(std::move(arr)) : mk_val<value_array>(std::move(arr));
         }},
-        {"unique", [](const func_args &) -> value {
-            throw not_implemented_exception("Array unique builtin not implemented");
-        }},
+        {"unique", array_unique_not_implemented},
     };
     return builtins;
 }
 
+[[noreturn]] static value object_join_not_implemented(const func_args &) {
+    throw not_implemented_exception("object join not implemented");
+}
 
 const func_builtins & value_object_t::get_builtins() const {
     if (!has_builtins) {
@@ -1183,9 +1189,7 @@ const func_builtins & value_object_t::get_builtins() const {
             });
             return result;
         }},
-        {"join", [](const func_args &) -> value {
-            throw not_implemented_exception("object join not implemented");
-        }},
+        {"join", object_join_not_implemented},
     };
     return builtins;
 }

common/jinja/value.h

@@ -129,27 +129,25 @@ struct value_t {
     // Note: only for debugging and error reporting purposes
     virtual std::string type() const { return ""; }
 
-    virtual int64_t as_int() const { throw std::runtime_error(type() + " is not an int value"); }
-    virtual double as_float() const { throw std::runtime_error(type() + " is not a float value"); }
-    virtual string as_string() const { throw std::runtime_error(type() + " is not a string value"); }
-    virtual bool as_bool() const { throw std::runtime_error(type() + " is not a bool value"); }
-    virtual const std::vector<value> & as_array() const { throw std::runtime_error(type() + " is not an array value"); }
-    virtual const std::vector<std::pair<value, value>> & as_ordered_object() const { throw std::runtime_error(type() + " is not an object value"); }
-    virtual value invoke(const func_args &) const { throw std::runtime_error(type() + " is not a function value"); }
+    virtual int64_t as_int() const { throw_type_error("is not an int value"); }
+    virtual double as_float() const { throw_type_error("is not a float value"); }
+    virtual string as_string() const { throw_type_error("is not a string value"); }
+    virtual bool as_bool() const { throw_type_error("is not a bool value"); }
+    virtual const std::vector<value> & as_array() const { throw_type_error("is not an array value"); }
+    virtual const std::vector<std::pair<value, value>> & as_ordered_object() const { throw_type_error("is not an object value"); }
+    virtual value invoke(const func_args &) const { throw_type_error("is not a function value"); }
     virtual bool is_none() const { return false; }
     virtual bool is_undefined() const { return false; }
-    virtual const func_builtins & get_builtins() const {
-        throw std::runtime_error("No builtins available for type " + type());
-    }
+    virtual const func_builtins & get_builtins() const { throw_type_error("has no builtins"); }
 
-    virtual bool has_key(const value &) { throw std::runtime_error(type() + " is not an object value"); }
-    virtual void insert(const value & /* key */, const value & /* val */) { throw std::runtime_error(type() + " is not an object value"); }
-    virtual value & at(const value & /* key */, value & /* default_val */) { throw std::runtime_error(type() + " is not an object value"); }
-    virtual value & at(const value & /* key */) { throw std::runtime_error(type() + " is not an object value"); }
-    virtual value & at(const std::string & /* key */, value & /* default_val */) { throw std::runtime_error(type() + " is not an object value"); }
-    virtual value & at(const std::string & /* key */) { throw std::runtime_error(type() + " is not an object value"); }
-    virtual value & at(int64_t /* idx */, value & /* default_val */) { throw std::runtime_error(type() + " is not an array value"); }
-    virtual value & at(int64_t /* idx */) { throw std::runtime_error(type() + " is not an array value"); }
+    virtual bool has_key(const value &) { throw_type_error("is not an object value"); }
+    virtual void insert(const value & /* key */, const value & /* val */) { throw_type_error("is not an object value"); }
+    virtual value & at(const value & /* key */, value & /* default_val */) { throw_type_error("is not an object value"); }
+    virtual value & at(const value & /* key */) { throw_type_error("is not an object value"); }
+    virtual value & at(const std::string & /* key */, value & /* default_val */) { throw_type_error("is not an object value"); }
+    virtual value & at(const std::string & /* key */) { throw_type_error("is not an object value"); }
+    virtual value & at(int64_t /* idx */, value & /* default_val */) { throw_type_error("is not an array value"); }
+    virtual value & at(int64_t /* idx */) { throw_type_error("is not an array value"); }
 
     virtual bool is_numeric() const { return false; }
     virtual bool is_hashable() const { return false; }
@@ -163,6 +161,11 @@ struct value_t {
     // Note: only for debugging purposes
     virtual std::string as_repr() const { return as_string().str(); }
 
+private:
+    [[noreturn]] void throw_type_error(const char* expected) const {
+        throw std::runtime_error(type() + " " + expected);
+    }
+
 protected:
     virtual bool equivalent(const value_t &) const = 0;
     virtual bool nonequal(const value_t & other) const { return !equivalent(other); }

common/log.cpp

@@ -23,6 +23,10 @@
 
 int common_log_verbosity_thold = LOG_DEFAULT_LLAMA;
 
+int common_log_get_verbosity_thold(void) {
+    return common_log_verbosity_thold;
+}
+
 void common_log_set_verbosity_thold(int verbosity) {
     common_log_verbosity_thold = verbosity;
 }
@@ -45,7 +49,7 @@ enum common_log_col : int {
 };
 
 // disable colors by default
-static std::vector<const char *> g_col = {
+static const char* g_col[] = {
     "",
     "",
     "",
@@ -243,7 +247,6 @@ public:
 
             entries = std::move(new_entries);
         }
-
         cv.notify_one();
     }
 
@@ -261,7 +264,6 @@ public:
                 {
                     std::unique_lock<std::mutex> lock(mtx);
                     cv.wait(lock, [this]() { return head != tail; });
-
                     cur = entries[head];
 
                     head = (head + 1) % entries.size();
@@ -297,7 +299,6 @@ public:
 
                 tail = (tail + 1) % entries.size();
             }
-
             cv.notify_one();
         }
 
@@ -334,7 +335,7 @@ public:
             g_col[COMMON_LOG_COL_CYAN]    = LOG_COL_CYAN;
             g_col[COMMON_LOG_COL_WHITE]   = LOG_COL_WHITE;
         } else {
-            for (size_t i = 0; i < g_col.size(); i++) {
+            for (size_t i = 0; i < std::size(g_col); i++) {
                 g_col[i] = "";
             }
         }
@@ -364,14 +365,20 @@ struct common_log * common_log_init() {
 }
 
 struct common_log * common_log_main() {
-    static struct common_log log;
+    // We intentionally leak (i.e. do not delete) the logger singleton because
+    // common_log destructor called at DLL teardown phase will cause hanging on Windows.
+    // OS will release resources anyway so it should not be a significant issue,
+    // though this design may cause logs to be lost if not flushed before the program exits.
+    // Refer to https://github.com/ggml-org/llama.cpp/issues/22142 for details.
+    static struct common_log * log;
     static std::once_flag    init_flag;
     std::call_once(init_flag, [&]() {
+        log = new common_log;
         // Set default to auto-detect colors
-        log.set_colors(tty_can_use_colors());
+        log->set_colors(tty_can_use_colors());
     });
 
-    return &log;
+    return log;
 }
 
 void common_log_pause(struct common_log * log) {

common/log.h

@@ -38,7 +38,7 @@ enum log_colors {
 
 // needed by the LOG_TMPL macro to avoid computing log arguments if the verbosity lower
 // set via common_log_set_verbosity()
-extern int common_log_verbosity_thold;
+int  common_log_get_verbosity_thold(void);
 
 void common_log_set_verbosity_thold(int verbosity); // not thread-safe
 
@@ -49,7 +49,11 @@ void common_log_default_callback(enum ggml_log_level level, const char * text, v
 struct common_log;
 
 struct common_log * common_log_init();
-struct common_log * common_log_main(); // singleton, automatically destroys itself on exit
+
+// Singleton, intentionally leaked to avoid Windows teardown hangs.
+// Call common_log_flush() before exit if you want to ensure all logs are flushed.
+struct common_log * common_log_main();
+
 void                common_log_pause (struct common_log * log); // pause  the worker thread, not thread-safe
 void                common_log_resume(struct common_log * log); // resume the worker thread, not thread-safe
 void                common_log_free  (struct common_log * log);
@@ -98,7 +102,7 @@ void common_log_flush         (struct common_log * log);                    // f
 
 #define LOG_TMPL(level, verbosity, ...) \
     do { \
-        if ((verbosity) <= common_log_verbosity_thold) { \
+        if ((verbosity) <= common_log_get_verbosity_thold()) { \
             common_log_add(common_log_main(), (level), __VA_ARGS__); \
         } \
     } while (0)

common/ngram-map.cpp

@@ -208,7 +208,7 @@ void common_ngram_map_begin(
                 count_keys, count_keys_del, count_values_del, count_map_entries_upd);
     }
 
-    map.idx_last_check = (map.size_last_begin > 0) ? map.size_last_begin - 1 : 0;
+    map.idx_last_check = size_begin;
     map.size_last_begin = size_begin;
 }
 
@@ -231,7 +231,7 @@ void common_ngram_map_draft(common_ngram_map & map,
         GGML_ABORT("%s: cur_len exceeds UINT32_MAX: %zu", __func__, cur_len);
     }
 
-    if (map.idx_last_check  > cur_len) {
+    if (map.idx_last_check > cur_len) {
         // Should not happen because of common_ngram_map_begin().
         GGML_ABORT("%s: map.idx_last_check > cur_len: %zu > %zu", __func__, map.idx_last_check, cur_len);
     }
@@ -386,7 +386,7 @@ void common_ngram_map_draft(common_ngram_map & map,
         LOG_DBG("%s: key_idx = %zu, key_offset = %zu, key_num = %d, draft.size = %zu\n", __func__,
                 curr_key.key_idx, key_offset, curr_key.key_num, draft.size());
 
-        map.last_draft_created   = false;
+        map.last_draft_created   = true;
         map.last_draft_key_idx   = key_offset;
         map.last_draft_value_idx = 0; // value 0 is used for simple mode
         return;
@@ -524,7 +524,7 @@ void common_ngram_map_accept(common_ngram_map & map, uint16_t n_accepted) {
     struct common_ngram_map_value & curr_value = curr_key.values[val_idx]; // value used for draft generation.
 
     // update the value statistics
-    LOG_INF("common_ngram_map_send_accepted: n_accepted = %d, prev value_num = %d\n",
+    LOG_DBG("common_ngram_map_send_accepted: n_accepted = %d, prev value_num = %d\n",
             n_accepted, curr_value.n_accepted);
     curr_value.n_accepted = n_accepted;
 }

common/preset.cpp

@@ -43,7 +43,7 @@ static std::set<std::string> get_remote_preset_whitelist(const std::map<std::str
     for (const auto & it : key_to_opt) {
         const std::string & key = it.first;
         const common_arg & opt = it.second;
-        if (allowed_options.find(key) != allowed_options.end() || opt.is_sparam) {
+        if (allowed_options.find(key) != allowed_options.end() || opt.is_sampling) {
             allowed_keys.insert(key);
             // also add variant keys (args without leading dashes and env vars)
             for (const auto & arg : opt.get_args()) {

common/reasoning-budget.cpp

@@ -122,6 +122,20 @@ static void common_reasoning_budget_accept(struct llama_sampler * smpl, llama_to
             }
             break;
         case REASONING_BUDGET_DONE:
+            // Re-arm on a new start tag: some models emit multiple <think> blocks
+            // per response, and each should get a fresh budget window.
+            if (ctx->start_matcher.advance(token)) {
+                ctx->state = REASONING_BUDGET_COUNTING;
+                ctx->remaining = ctx->budget;
+                ctx->end_matcher.reset();
+                LOG_INF("reasoning-budget: re-activated on new start tag, budget=%d tokens\n", ctx->budget);
+
+                if (ctx->remaining <= 0) {
+                    ctx->state = REASONING_BUDGET_FORCING;
+                    ctx->force_pos = 0;
+                    LOG_INF("reasoning-budget: budget=0, forcing immediately\n");
+                }
+            }
             break;
     }
 }
@@ -218,34 +232,6 @@ static struct llama_sampler * common_reasoning_budget_init_state(
     );
 }
 
-struct llama_sampler * common_reasoning_budget_init(
-        const struct llama_vocab       * vocab,
-        const std::vector<llama_token> & start_tokens,
-        const std::vector<llama_token> & end_tokens,
-        const std::vector<llama_token> & forced_tokens,
-        int32_t                          budget,
-        const std::vector<llama_token> & prefill_tokens) {
-    // Determine initial state from prefill: COUNTING if the prefill begins with
-    // the start sequence but does not also contain the end sequence after it.
-    common_reasoning_budget_state initial_state = REASONING_BUDGET_IDLE;
-    if (!prefill_tokens.empty() && !start_tokens.empty() &&
-            prefill_tokens.size() >= start_tokens.size() &&
-            std::equal(start_tokens.begin(), start_tokens.end(), prefill_tokens.begin())) {
-        initial_state = REASONING_BUDGET_COUNTING;
-        // If the end sequence also follows the start in the prefill, reasoning
-        // was opened and immediately closed — stay IDLE.
-        if (!end_tokens.empty() &&
-                prefill_tokens.size() >= start_tokens.size() + end_tokens.size()) {
-            auto end_start = prefill_tokens.end() - (ptrdiff_t) end_tokens.size();
-            if (end_start >= prefill_tokens.begin() + (ptrdiff_t) start_tokens.size() &&
-                    std::equal(end_tokens.begin(), end_tokens.end(), end_start)) {
-                initial_state = REASONING_BUDGET_IDLE;
-            }
-        }
-    }
-    return common_reasoning_budget_init_state(vocab, start_tokens, end_tokens, forced_tokens, budget, initial_state);
-}
-
 struct llama_sampler * common_reasoning_budget_init(
         const struct llama_vocab       * vocab,
         const std::vector<llama_token> & start_tokens,

common/reasoning-budget.h

@@ -29,10 +29,7 @@ enum common_reasoning_budget_state {
 //   end_tokens     - token sequence for natural deactivation
 //   forced_tokens  - token sequence forced when budget expires
 //   budget         - max tokens allowed in the reasoning block
-//   prefill_tokens - tokens already present in the prompt (generation prompt);
-//                    used to determine the initial state: COUNTING if they begin
-//                    with start_tokens (but don't also end with end_tokens),
-//                    IDLE otherwise. COUNTING with budget <= 0 is promoted to FORCING.
+//   initial_state  - initial state
 //
 struct llama_sampler * common_reasoning_budget_init(
         const struct llama_vocab       * vocab,
@@ -40,16 +37,6 @@ struct llama_sampler * common_reasoning_budget_init(
         const std::vector<llama_token> & end_tokens,
         const std::vector<llama_token> & forced_tokens,
         int32_t                          budget,
-        const std::vector<llama_token> & prefill_tokens = {});
-
-// Variant that takes an explicit initial state (used by tests and clone).
-// COUNTING with budget <= 0 is promoted to FORCING.
-struct llama_sampler * common_reasoning_budget_init(
-        const struct llama_vocab       * vocab,
-        const std::vector<llama_token> & start_tokens,
-        const std::vector<llama_token> & end_tokens,
-        const std::vector<llama_token> & forced_tokens,
-        int32_t                          budget,
-        common_reasoning_budget_state    initial_state);
+        common_reasoning_budget_state    initial_state = REASONING_BUDGET_IDLE);
 
 common_reasoning_budget_state common_reasoning_budget_get_state(const struct llama_sampler * smpl);

common/sampling.cpp

@@ -1,10 +1,12 @@
 #include "sampling.h"
 
 #include "common.h"
-#include "ggml.h"
+#include "fit.h"
 #include "log.h"
 #include "reasoning-budget.h"
 
+#include "ggml.h"
+
 #include <algorithm>
 #include <cctype>
 #include <climits>
@@ -258,32 +260,35 @@ struct common_sampler * common_sampler_init(const struct llama_model * model, st
         }
     }
 
+    // Compute prefill tokens from the generation prompt
+    std::vector<llama_token> prefill_tokens;
+    if (!params.generation_prompt.empty()) {
+        GGML_ASSERT(vocab != nullptr);
+        auto tokens = common_tokenize(vocab, params.generation_prompt, false, true);
+        for (size_t i = 0; i < tokens.size(); i++) {
+            std::string piece = common_token_to_piece(vocab, tokens[i], true);
+            if (i == 0 && std::isspace(piece[0]) && !std::isspace(params.generation_prompt[0])) {
+                // Some tokenizers will add a space before the first special token, need to exclude
+                continue;
+            }
+            LOG_DBG("%s: prefill token: %d = %s\n", __func__, tokens[i], piece.c_str());
+            prefill_tokens.push_back(tokens[i]);
+        }
+    }
+
     // Feed generation prompt tokens to the grammar sampler so it advances past
     // tokens the template already placed in the prompt.
     // Only applies to output-format and tool-call grammars; user-supplied grammars must not be prefilled.
-    std::vector<llama_token> prefill_tokens;
-    if (!params.generation_prompt.empty() && common_grammar_needs_prefill(params.grammar)) {
-        GGML_ASSERT(vocab != nullptr);
-        prefill_tokens = common_tokenize(vocab, params.generation_prompt, false, true);
-        if (!prefill_tokens.empty()) {
-            std::string first_token = common_token_to_piece(vocab, prefill_tokens[0], true);
-            if (std::isspace(first_token[0]) && !std::isspace(params.generation_prompt[0])) {
-                // Some tokenizers will add a space before the first special token, need to remove
-                prefill_tokens = std::vector<llama_token>(prefill_tokens.begin() + 1, prefill_tokens.end());
-            }
-        }
-
-        if (grmr && !params.grammar_lazy) {
-            try {
-                for (const auto & token : prefill_tokens) {
-                    llama_sampler_accept(grmr, token);
-                    LOG_DBG("%s: accepted prefill token (%d)\n", __func__, token);
-                }
-            } catch (std::exception &e) {
-                LOG_ERR("%s: error initializing grammar sampler for grammar:\n%s\n\nGeneration prompt:\n'%s'\n", __func__,
-                    common_grammar_value(params.grammar).c_str(), params.generation_prompt.c_str());
-                throw e;
+    if (grmr && !params.grammar_lazy && common_grammar_needs_prefill(params.grammar)) {
+        try {
+            for (const auto & token : prefill_tokens) {
+                llama_sampler_accept(grmr, token);
+                LOG_DBG("%s: grammar accepted prefill token (%d)\n", __func__, token);
             }
+        } catch (std::exception &e) {
+            LOG_ERR("%s: error initializing grammar sampler for grammar:\n%s\n\nGeneration prompt:\n'%s'\n", __func__,
+                common_grammar_value(params.grammar).c_str(), params.generation_prompt.c_str());
+            throw e;
         }
     }
 
@@ -294,8 +299,12 @@ struct common_sampler * common_sampler_init(const struct llama_model * model, st
             params.reasoning_budget_start,
             params.reasoning_budget_end,
             params.reasoning_budget_forced,
-            params.reasoning_budget_tokens < 0 ? INT_MAX : params.reasoning_budget_tokens,
-            prefill_tokens);
+            params.reasoning_budget_tokens < 0 ? INT_MAX : params.reasoning_budget_tokens);
+
+        for (const auto & token : prefill_tokens) {
+            llama_sampler_accept(rbudget, token);
+            LOG_DBG("%s: reasoning-budget accepted prefill token (%d)\n", __func__, token);
+        }
     }
 
     if (params.has_logit_bias()) {
@@ -429,7 +438,7 @@ static bool grammar_should_apply(struct common_sampler * gsmpl) {
     return true;
 }
 
-void common_sampler_accept(struct common_sampler * gsmpl, llama_token token, bool accept_grammar) {
+void common_sampler_accept(struct common_sampler * gsmpl, llama_token token, bool is_generated) {
     if (!gsmpl) {
         return;
     }
@@ -437,9 +446,11 @@ void common_sampler_accept(struct common_sampler * gsmpl, llama_token token, boo
     const auto tm = gsmpl->tm();
 
     // grammar_should_apply() checks the reasoning budget state, so calculate this before we accept
-    accept_grammar = accept_grammar && grammar_should_apply(gsmpl);
+    const auto accept_grammar = is_generated && grammar_should_apply(gsmpl);
 
-    llama_sampler_accept(gsmpl->rbudget, token);
+    if (gsmpl->rbudget && is_generated) {
+        llama_sampler_accept(gsmpl->rbudget, token);
+    }
 
     if (gsmpl->grmr && accept_grammar) {
         llama_sampler_accept(gsmpl->grmr, token);
@@ -511,7 +522,7 @@ void common_perf_print(const struct llama_context * ctx, const struct common_sam
         LOG_INF("%s: unaccounted time = %10.2f ms / %5.1f %%      (total - sampling - prompt eval - eval) / (total)\n", __func__, t_unacc_ms, t_unacc_pc);
         LOG_INF("%s:    graphs reused = %10d\n", __func__, data.n_reused);
 
-        llama_memory_breakdown_print(ctx);
+        common_memory_breakdown_print(ctx);
     }
 }
 

common/sampling.h

@@ -41,8 +41,8 @@ struct common_sampler * common_sampler_init(const struct llama_model * model, st
 
 void common_sampler_free(struct common_sampler * gsmpl);
 
-// if accept_grammar is true, the token is accepted both by the sampling chain and the grammar
-void                    common_sampler_accept(struct common_sampler * gsmpl, llama_token token, bool accept_grammar);
+// if is_generated is true, the token is accepted by the sampling chain, the reasoning budget sampler, and the grammar sampler
+void                    common_sampler_accept(struct common_sampler * gsmpl, llama_token token, bool is_generated);
 void                    common_sampler_reset (struct common_sampler * gsmpl);
 struct common_sampler * common_sampler_clone (struct common_sampler * gsmpl);
 

common/speculative.cpp

@@ -13,6 +13,7 @@
 #include <cstring>
 #include <iomanip>
 #include <map>
+#include <cinttypes>
 
 #define SPEC_VOCAB_MAX_SIZE_DIFFERENCE  128
 #define SPEC_VOCAB_CHECK_START_TOKEN_ID 5
@@ -60,18 +61,26 @@ static bool common_speculative_are_compatible(
     LOG_DBG("%s: vocab_type dft: %d\n", __func__, vocab_type_dft);
 
     if (vocab_type_tgt != vocab_type_dft) {
-        LOG_DBG("%s: draft model vocab type must match target model to use speculation but ", __func__);
-        LOG_DBG("vocab_type_dft = %d while vocab_type_tgt = %d\n", vocab_type_dft, vocab_type_tgt);
+        LOG_WRN("%s: draft model vocab type must match target model to use speculation but "
+                "vocab_type_dft = %d while vocab_type_tgt = %d\n", __func__, vocab_type_dft, vocab_type_tgt);
         return false;
     }
 
-    if (
-        llama_vocab_get_add_bos(vocab_tgt) != llama_vocab_get_add_bos(vocab_dft) ||
-        llama_vocab_get_add_eos(vocab_tgt) != llama_vocab_get_add_eos(vocab_dft) ||
-        llama_vocab_bos(vocab_tgt) != llama_vocab_bos(vocab_dft) ||
-        llama_vocab_eos(vocab_tgt) != llama_vocab_eos(vocab_dft)
-    ) {
-        LOG_DBG("%s: draft model special tokens must match target model to use speculation\n", __func__);
+    if (llama_vocab_get_add_bos(vocab_tgt) != llama_vocab_get_add_bos(vocab_dft) ||
+        (llama_vocab_get_add_bos(vocab_tgt) && llama_vocab_bos(vocab_tgt) != llama_vocab_bos(vocab_dft))) {
+        LOG_WRN("%s: draft model bos tokens must match target model to use speculation. add: %d - %d, id: %d - %d)\n",
+                __func__,
+                llama_vocab_get_add_bos(vocab_tgt), llama_vocab_get_add_bos(vocab_dft),
+                llama_vocab_bos(vocab_tgt), llama_vocab_bos(vocab_dft));
+        return false;
+    }
+
+    if (llama_vocab_get_add_eos(vocab_tgt) != llama_vocab_get_add_eos(vocab_dft) ||
+        (llama_vocab_get_add_eos(vocab_tgt) && llama_vocab_eos(vocab_tgt) != llama_vocab_eos(vocab_dft))) {
+        LOG_WRN("%s: draft model eos tokens must match target model to use speculation. add: %d - %d, id: %d - %d)\n",
+                __func__,
+                llama_vocab_get_add_eos(vocab_tgt), llama_vocab_get_add_eos(vocab_dft),
+                llama_vocab_eos(vocab_tgt), llama_vocab_eos(vocab_dft));
         return false;
     }
 
@@ -142,12 +151,31 @@ struct common_speculative_state {
             llama_tokens & result) = 0;
 
     virtual void accept(uint16_t n_accepted) = 0;
+
+    virtual int32_t n_max(const common_params_speculative & params) const = 0;
+    virtual int32_t n_min(const common_params_speculative & params) const = 0;
+};
+
+struct common_speculative_checkpoint {
+    llama_pos pos_min  = 0;
+    llama_pos pos_max  = 0;
+
+    int64_t   n_tokens = 0;
+
+    std::vector<uint8_t> data;
+
+    size_t size() const {
+        return data.size();
+    }
 };
 
 struct common_speculative_state_draft : public common_speculative_state {
     llama_context * ctx_tgt; // only used for retokenizing from ctx_dft
     llama_context * ctx_dft;
 
+    bool use_ckpt = false;
+    common_speculative_checkpoint ckpt;
+
     common_sampler * smpl;
 
     llama_batch  batch;
@@ -160,10 +188,12 @@ struct common_speculative_state_draft : public common_speculative_state {
             enum common_speculative_type type,
             llama_context * ctx_tgt,
             llama_context * ctx_dft,
-            const std::vector<std::pair<std::string, std::string>> & replacements)
+            const std::vector<std::pair<std::string, std::string>> & replacements,
+            bool use_ckpt)
         : common_speculative_state(type)
         , ctx_tgt(ctx_tgt)
         , ctx_dft(ctx_dft)
+        , use_ckpt(use_ckpt)
     {
         batch = llama_batch_init(llama_n_batch(ctx_dft), 0, 1);
         smpl = nullptr;
@@ -217,8 +247,39 @@ struct common_speculative_state_draft : public common_speculative_state {
         llama_batch_free(batch);
     }
 
-    void begin(const llama_tokens & prompt) override {
-        GGML_UNUSED(prompt);
+    void begin(const llama_tokens & /*prompt*/) override {
+    }
+
+    size_t create_checkpoint(int n_tokens_prompt) {
+        int slot_id = 0;
+        const size_t checkpoint_size = llama_state_seq_get_size_ext(ctx_dft, slot_id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);
+
+        ckpt.pos_min  = llama_memory_seq_pos_min(llama_get_memory(ctx_dft), slot_id);
+        ckpt.pos_max  = llama_memory_seq_pos_max(llama_get_memory(ctx_dft), slot_id);
+        ckpt.n_tokens = n_tokens_prompt;
+        ckpt.data.resize(checkpoint_size);
+
+        const size_t n = llama_state_seq_get_data_ext(ctx_dft, ckpt.data.data(), checkpoint_size, slot_id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);
+        if (n != checkpoint_size) {
+            GGML_ABORT("checkpoint size mismatch: expected %zu, got %zu\n", checkpoint_size, n);
+        }
+
+        LOG_DBG("%s: pos_min = %d, pos_max = %d, size = %.3f MiB\n", __func__,
+                ckpt.pos_min, ckpt.pos_max, (float) ckpt.data.size() / 1024 / 1024);
+        return n;
+    }
+
+    size_t restore_checkpoint() {
+        int slot_id = 0;
+        LOG_DBG("%s: pos_min = %d, pos_max = %d\n", __func__, ckpt.pos_min, ckpt.pos_max);
+        const size_t n = llama_state_seq_set_data_ext(ctx_dft, ckpt.data.data(), ckpt.size(), slot_id, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);
+        if (n != ckpt.size()) {
+            GGML_ABORT("%s: failed to restore context checkpoint (pos_min=%d, pos_max=%d, size=%zu",
+                        __func__, ckpt.pos_min, ckpt.pos_max, ckpt.size());
+        }
+        llama_memory_seq_rm(llama_get_memory(ctx_dft), slot_id, ckpt.pos_max + 1, -1);
+
+        return n;
     }
 
     void draft(
@@ -226,6 +287,8 @@ struct common_speculative_state_draft : public common_speculative_state {
             const llama_tokens & prompt_tgt,
             llama_token id_last,
             llama_tokens & result) override {
+        const auto & sparams = params.draft;
+
         auto * spec = this;
 
         auto & batch      = spec->batch;
@@ -236,10 +299,10 @@ struct common_speculative_state_draft : public common_speculative_state {
 
         auto * mem_dft = llama_get_memory(ctx_dft);
 
-        int reuse_i = 0;
-        int reuse_n = 0;
+        int reuse_i = 0; // index of part to be reused in prompt_dft
+        int reuse_n = 0; // length of part to be reused in prompt_dft
 
-        const int n_ctx = llama_n_ctx(ctx_dft) - params.n_max;
+        const int n_ctx = llama_n_ctx(ctx_dft) - sparams.n_max;
 
         llama_tokens prompt_cnv;
         if (!spec->vocab_cmpt) {
@@ -271,13 +334,18 @@ struct common_speculative_state_draft : public common_speculative_state {
 
         const int i_start = std::max<int>(0, (int) prompt_cur.size() - n_ctx);
 
+        if (use_ckpt && i_start > 0) {
+            LOG_WRN("%s: context shift is not supported with checkpoint-based contexts - skipping\n", __func__);
+            return;
+        }
+
         // reuse as much as possible from the old draft context
         // ideally, the draft context should be as big as the target context and we will always reuse the entire prompt
         for (int i = 0; i < (int) prompt_dft.size(); ++i) {
             int cur = 0;
             while (i_start + cur < (int) prompt_cur.size() &&
-                    i       + cur < (int) prompt_dft.size() &&
-                    prompt_cur[i_start + cur] == prompt_dft[i + cur]) {
+                   i       + cur < (int) prompt_dft.size() &&
+                   prompt_cur[i_start + cur] == prompt_dft[i + cur]) {
                 cur++;
             }
 
@@ -285,24 +353,37 @@ struct common_speculative_state_draft : public common_speculative_state {
                 reuse_i = i;
                 reuse_n = cur;
             }
+
+            if (use_ckpt) {
+                break;
+            }
         }
 
-        LOG_DBG("%s: reuse_i = %d, reuse_n = %d, prompt = %d\n", __func__, reuse_i, reuse_n, (int) prompt_dft.size());
+        LOG_DBG("%s: reuse_i = %d, reuse_n = %d, #prompt_dft = %zu, #prompt_cur = %zu\n",
+                __func__, reuse_i, reuse_n, prompt_dft.size(), prompt_cur.size());
+        if (use_ckpt && ckpt.n_tokens > reuse_n) {
+            LOG_DBG("%s: checkpoint (n_tokens = %d) is outdated -> delete it\n", __func__, (int) ckpt.n_tokens);
+
+            reuse_i = 0;
+            reuse_n = 0;
+
+            ckpt = {};
+        }
 
         result.clear();
-        result.reserve(params.n_max);
+        result.reserve(sparams.n_max);
 
-        if (reuse_n == 0) {
+        if (reuse_n == 0 || (use_ckpt && reuse_i > 0)) {
             llama_memory_clear(mem_dft, false);
             prompt_dft.clear();
         } else {
             // this happens when a previous draft has been discarded (for example, due to being too small), but the
             // target model agreed with it. in this case, we simply pass back the previous results to save compute
-            if (reuse_i + reuse_n < (int) prompt_dft.size() && prompt_dft[reuse_i + reuse_n] == id_last) {
+            if (reuse_i + reuse_n < (int64_t) prompt_dft.size() && prompt_dft[reuse_i + reuse_n] == id_last) {
                 for (int i = reuse_i + reuse_n + 1; i < (int) prompt_dft.size(); ++i) {
                     result.push_back(prompt_dft[i]);
 
-                    if (params.n_max <= (int) result.size()) {
+                    if (sparams.n_max <= (int) result.size()) {
                         break;
                     }
                 }
@@ -311,15 +392,34 @@ struct common_speculative_state_draft : public common_speculative_state {
             }
 
             if (reuse_i > 0) {
-                llama_memory_seq_rm (mem_dft, 0, 0, reuse_i);
+                GGML_ASSERT(!use_ckpt);
+
+                bool is_removed = llama_memory_seq_rm (mem_dft, 0, 0, reuse_i);
+                if (!is_removed) {
+                    LOG_ERR("%s: llama_memory_seq_rm failed, reuse_i=%d\n", __func__, reuse_i);
+                    return;
+                }
                 llama_memory_seq_add(mem_dft, 0, reuse_i, -1, -reuse_i);
 
                 prompt_dft.erase(prompt_dft.begin(), prompt_dft.begin() + reuse_i);
             }
 
             if (reuse_n < (int) prompt_dft.size()) {
-                llama_memory_seq_rm (mem_dft, 0, reuse_n, -1);
-                prompt_dft.erase(prompt_dft.begin() + reuse_n, prompt_dft.end());
+                if (use_ckpt) {
+                    if (ckpt.n_tokens > 0) {
+                        LOG_DBG("%s: restoring checkpoint, reuse_n=%d, prompt_dft.size=%zu\n", __func__, reuse_n, prompt_dft.size());
+                        restore_checkpoint();
+                        reuse_n = ckpt.n_tokens;
+                        prompt_dft.resize(reuse_n);
+                    }
+                } else {
+                    const bool is_removed = llama_memory_seq_rm(mem_dft, 0, reuse_n, -1);
+                    if (!is_removed) {
+                        LOG_ERR("%s: llama_memory_seq_rm failed, reuse_n=%d, prompt_dft.size=%zu\n", __func__, reuse_n, prompt_dft.size());
+                        return;
+                    }
+                    prompt_dft.erase(prompt_dft.begin() + reuse_n, prompt_dft.end());
+                }
             }
         }
 
@@ -336,8 +436,17 @@ struct common_speculative_state_draft : public common_speculative_state {
         // we should rarely end-up here during normal decoding
         if (batch.n_tokens > 0) {
             //LOG_DBG("%s: draft prompt batch: %s\n", __func__, string_from(ctx, batch).c_str());
+            LOG_DBG("%s: draft prompt batch: %d tokens\n", __func__, batch.n_tokens);
 
-            llama_decode(ctx_dft, batch);
+            int ret = llama_decode(ctx_dft, batch);
+            if (ret != 0 && ret != 1) {
+                LOG_WRN("%s: llama_decode returned %d, prompt_cur.size=%zu\n",
+                        __func__, ret, prompt_cur.size());
+            }
+
+            if (use_ckpt) {
+                create_checkpoint(prompt_dft.size());
+            }
         }
 
         const llama_pos n_past = prompt_dft.size();
@@ -349,14 +458,18 @@ struct common_speculative_state_draft : public common_speculative_state {
 
         prompt_dft.push_back(id_last);
 
-        LOG_DBG("%s: draft prompt: %s\n", __func__, string_from(ctx_dft, prompt_dft).c_str());
+        //LOG_DBG("%s: draft prompt: %s\n", __func__, string_from(ctx_dft, prompt_dft).c_str());
 
-        llama_decode(ctx_dft, batch);
+        int ret = llama_decode(ctx_dft, batch);
+        if (ret != 0 && ret != 1) {
+            LOG_WRN("%s: llama_decode returned %d, prompt_cur.size=%zu, prompt_dft.size=%zu\n",
+                    __func__, ret, prompt_cur.size(), prompt_dft.size());
+        }
 
         common_sampler_reset(smpl);
 
         // sample n_draft tokens from the draft model
-        for (int i = 0; i < params.n_max; ++i) {
+        for (int i = 0; i < sparams.n_max; ++i) {
             common_batch_clear(batch);
 
             common_sampler_sample(smpl, ctx_dft, 0, true);
@@ -373,21 +486,25 @@ struct common_speculative_state_draft : public common_speculative_state {
 
             common_sampler_accept(smpl, id, true);
 
-            result.push_back(id);
-
-            if (params.n_max <= (int) result.size()) {
+            // only collect very high-confidence draft tokens
+            if (cur_p->data[0].p < sparams.p_min) {
                 break;
             }
 
-            // only collect very high-confidence draft tokens
-            if (cur_p->data[0].p < params.p_min) {
+            result.push_back(id);
+
+            if (sparams.n_max <= (int) result.size()) {
                 break;
             }
 
             common_batch_add(batch, id, n_past + i + 1, { 0 }, true);
 
             // evaluate the drafted tokens on the draft model
-            llama_decode(ctx_dft, batch);
+            ret = llama_decode(ctx_dft, batch);
+            if (ret != 0) {
+                LOG_WRN("%s: llama_decode[%d] returned %d, prompt_cur.size=%zu, prompt_dft.size=%zu\n",
+                        __func__, i, ret, prompt_cur.size(), prompt_dft.size());
+            }
 
             prompt_dft.push_back(id);
         }
@@ -397,10 +514,14 @@ struct common_speculative_state_draft : public common_speculative_state {
             detokenized = replace_to_tgt(detokenized);
             LOG_DBG("draft->main detokenized string: '%s'\n", detokenized.c_str());
             result = common_tokenize(ctx_tgt, detokenized, false, true);
-            if (result.size() > (size_t)params.n_max) {
-                result.resize(params.n_max);
+            if (result.size() > (size_t) sparams.n_max) {
+                result.resize(sparams.n_max);
             }
         }
+
+        if (result.size() < (size_t) sparams.n_min) {
+            result.clear();
+        }
     }
 
     void accept(uint16_t n_accepted) override {
@@ -408,6 +529,14 @@ struct common_speculative_state_draft : public common_speculative_state {
         GGML_UNUSED(n_accepted);
     }
 
+    int32_t n_max(const common_params_speculative & params) const override {
+        return params.draft.n_max;
+    }
+
+    int32_t n_min(const common_params_speculative & params) const override {
+        return params.draft.n_min;
+    }
+
     std::string replace_to_dft(const std::string & input) const {
         std::string result = input;
 
@@ -460,6 +589,14 @@ struct common_speculative_state_eagle3 : public common_speculative_state {
         // noop
         GGML_UNUSED(n_accepted);
     }
+
+    int32_t n_max(const common_params_speculative & params) const override {
+        return params.draft.n_max;
+    }
+
+    int32_t n_min(const common_params_speculative & params) const override {
+        return params.draft.n_min;
+    }
 };
 
 // state of self-speculation (simple implementation, not ngram-map)
@@ -489,19 +626,27 @@ struct common_speculative_state_ngram_simple : public common_speculative_state {
         // noop
         GGML_UNUSED(n_accepted);
     }
+
+    int32_t n_max(const common_params_speculative & /*params*/) const override {
+        return config.size_mgram;
+    }
+
+    int32_t n_min(const common_params_speculative & /*params*/) const override {
+        return config.size_mgram;
+    }
 };
 
 struct common_speculative_state_ngram_map_k : public common_speculative_state {
     // draft ngram map for speculative decoding without draft model
-    common_ngram_map map;
+    common_ngram_map config;
 
     common_speculative_state_ngram_map_k(
             enum common_speculative_type type,
-            common_ngram_map map)
-        : common_speculative_state(type), map(std::move(map)) {}
+            common_ngram_map config)
+        : common_speculative_state(type), config(std::move(config)) {}
 
     void begin(const llama_tokens & prompt) override {
-        common_ngram_map_begin(map, prompt);
+        common_ngram_map_begin(config, prompt);
     }
 
     void draft(
@@ -509,12 +654,20 @@ struct common_speculative_state_ngram_map_k : public common_speculative_state {
             const llama_tokens & prompt_tgt,
             llama_token id_last,
             llama_tokens & result) override {
-        common_ngram_map_draft(map, prompt_tgt, id_last, result);
+        common_ngram_map_draft(config, prompt_tgt, id_last, result);
         GGML_UNUSED(params);
     }
 
     void accept(uint16_t n_accepted) override {
-        common_ngram_map_accept(map, n_accepted);
+        common_ngram_map_accept(config, n_accepted);
+    }
+
+    int32_t n_max(const common_params_speculative & /*params*/) const override {
+        return config.size_value;
+    }
+
+    int32_t n_min(const common_params_speculative & /*params*/) const override {
+        return config.size_value;
     }
 };
 
@@ -571,7 +724,7 @@ struct common_speculative_state_ngram_mod : public common_speculative_state {
             const llama_tokens & prompt_tgt,
             llama_token id_last,
             llama_tokens & result) override {
-        GGML_UNUSED(params);
+        const auto & sparams = params.ngram_mod;
 
         n_draft_last = 0;
 
@@ -591,16 +744,16 @@ struct common_speculative_state_ngram_mod : public common_speculative_state {
             i_last = cur_len - n;
         }
 
-        result.resize(n + params.n_max);
+        result.resize(n + sparams.n_max);
         for (size_t i = 0; i < n - 1; ++i) {
             result[i] = prompt_tgt[cur_len - n + 1 + i];
         }
         result[n - 1] = id_last;
 
-        for (int i = 0; i < params.n_max; ++i) {
+        for (int i = 0; i < sparams.n_max; ++i) {
             const llama_token token = mod.get(result.data() + i);
             if (token == common_ngram_mod::EMPTY) {
-                if (i < params.n_min) {
+                if (i < sparams.n_min) {
                     result.clear();
                     return;
                 }
@@ -622,26 +775,33 @@ struct common_speculative_state_ngram_mod : public common_speculative_state {
     }
 
     void accept(uint16_t n_accepted) override {
-        if (verbose) {
-            LOG_INF("%s: accepted %d tokens from %zu drafted tokens\n", __func__, n_accepted, n_draft_last);
-        }
-
         // compute acceptance fraction if we have a recorded draft length
         if (n_draft_last > 0) {
             const double f_acc = (double)n_accepted / (double)n_draft_last;
             if (f_acc < 0.5) {
                 n_low++;
                 if (n_low >= 3) {
-                    LOG_WRN("%s: low acceptance streak (%d) – resetting ngram_mod\n", __func__, n_low);
+                    if (verbose) {
+                        LOG_WRN("%s: low acceptance streak (%d) – resetting ngram_mod\n", __func__, n_low);
+                    }
 
                     mod.reset();
                     n_low = 0;
+                    i_last = 0;
                 }
             } else {
                 n_low = 0;
             }
         }
     }
+
+    int32_t n_max(const common_params_speculative & params) const override {
+        return params.ngram_mod.n_max;
+    }
+
+    int32_t n_min(const common_params_speculative & params) const override {
+        return params.ngram_mod.n_min;
+    }
 };
 
 struct common_speculative_state_ngram_cache : public common_speculative_state {
@@ -735,18 +895,29 @@ struct common_speculative_state_ngram_cache : public common_speculative_state {
         // TODO: noop
         GGML_UNUSED(n_accepted);
     }
+
+    int32_t n_max(const common_params_speculative & /*params*/) const override {
+        return n_draft;
+    }
+
+    int32_t n_min(const common_params_speculative & /*params*/) const override {
+        return 0;
+    }
 };
 
 struct common_speculative {
     std::vector<std::unique_ptr<common_speculative_state>> impls; // list of implementations to use and their states
+
     common_speculative_state * curr_impl = nullptr; // current implementation in use (for stats)
 };
 
-static common_ngram_map get_common_ngram_map(const common_speculative_config & config) {
-    uint16_t size_key   = config.params.ngram_size_n;
-    uint16_t size_value = config.params.ngram_size_m;
-    bool     key_only   = (config.type == COMMON_SPECULATIVE_TYPE_NGRAM_MAP_K);
-    uint16_t min_hits   = config.params.ngram_min_hits;
+static common_ngram_map get_common_ngram_map(
+        common_speculative_type type,
+        const common_params_speculative_ngram_map & config) {
+    uint16_t size_key   = config.size_n;
+    uint16_t size_value = config.size_m;
+    bool     key_only   = type == COMMON_SPECULATIVE_TYPE_NGRAM_MAP_K;
+    uint16_t min_hits   = config.min_hits;
 
     return common_ngram_map(size_key, size_value, key_only, min_hits);
 }
@@ -798,50 +969,14 @@ enum common_speculative_type common_speculative_type_from_name(const std::string
     return it->second;
 }
 
-bool common_speculative_is_compat(llama_context * ctx_tgt) {
-    auto * mem = llama_get_memory(ctx_tgt);
-    if (mem == nullptr) {
-        return false;
-    }
-
-    bool res = true;
-
-    llama_memory_clear(mem, true);
-
-    // eval 2 tokens to check if the context is compatible
-    std::vector<llama_token> tmp;
-    tmp.push_back(0);
-    tmp.push_back(0);
-
-    int ret = llama_decode(ctx_tgt, llama_batch_get_one(tmp.data(), tmp.size()));
-    if (ret != 0) {
-        LOG_ERR("%s: llama_decode() failed: %d\n", __func__, ret);
-        res = false;
-        goto done;
-    }
-
-    // try to remove the last tokens
-    if (!llama_memory_seq_rm(mem, 0, 1, -1)) {
-        LOG_WRN("%s: the target context does not support partial sequence removal\n", __func__);
-        res = false;
-        goto done;
-    }
-
-done:
-    llama_memory_clear(mem, true);
-    llama_synchronize(ctx_tgt);
-
-    return res;
-}
-
 // initialization of the speculative decoding system
 //
 common_speculative * common_speculative_init(
         common_params_speculative & params,
         llama_context             * ctx_tgt) {
     llama_context * ctx_dft = nullptr;
-    if (params.model_dft) {
-        ctx_dft = llama_init_from_model(params.model_dft, params.cparams_dft);
+    if (params.draft.model) {
+        ctx_dft = llama_init_from_model(params.draft.model, params.draft.cparams);
         if (ctx_dft == nullptr) {
             LOG_ERR("%s", "failed to create draft context\n");
             return nullptr;
@@ -851,7 +986,7 @@ common_speculative * common_speculative_init(
     // Compute the implementations to use based on the config and their order of preference
     std::vector<common_speculative_config> configs = {}; // list of speculative configs to try
     {
-        bool has_draft = !params.mparams_dft.path.empty();
+        bool has_draft = !params.draft.mparams.path.empty();
         bool has_draft_eagle3 = false; // TODO PR-18039: if params.speculative.eagle3
 
         bool has_ngram_cache   = (params.type == COMMON_SPECULATIVE_TYPE_NGRAM_CACHE);
@@ -874,16 +1009,17 @@ common_speculative * common_speculative_init(
             configs.push_back(common_speculative_config(COMMON_SPECULATIVE_TYPE_NGRAM_MAP_K4V, params));
         }
         if (has_ngram_mod) {
-            // shared instance for all speculative decoding contexts
-            if (!params.ngram_mod) {
-                params.ngram_mod = std::make_shared<common_ngram_mod>(params.ngram_size_n, 4*1024*1024);
+            auto & sparams = params.ngram_mod;
 
-                LOG_INF("%s: initialized ngram_mod with n=%d, size=%zu (%.3f MB)\n", __func__,
-                        params.ngram_size_n, params.ngram_mod->size(),
-                        (float)(params.ngram_mod->size_bytes())/1024/1024);
+            if (!sparams.obj) {
+                sparams.obj = std::make_shared<common_ngram_mod>(sparams.n_match, 4*1024*1024);
 
-                if (params.ngram_size_n < 16) {
-                    LOG_WRN("%s: ngram_mod n=%d is too small - poor quality is possible, see: https://github.com/ggml-org/llama.cpp/pull/19164\n", __func__, params.ngram_size_n);
+                LOG_INF("%s: initialized ngram_mod with n_match=%d, size=%zu (%.3f MB)\n", __func__,
+                        sparams.n_match, sparams.obj->size(), (float)(sparams.obj->size_bytes())/1024/1024);
+
+                if (sparams.n_match < 16) {
+                    LOG_WRN("%s: ngram_mod n_match=%d is too small - poor quality is possible, "
+                            "see: https://github.com/ggml-org/llama.cpp/pull/19164\n", __func__, sparams.n_match);
                 }
             }
 
@@ -908,10 +1044,13 @@ common_speculative * common_speculative_init(
             case COMMON_SPECULATIVE_TYPE_NONE:
                 break;
             case COMMON_SPECULATIVE_TYPE_DRAFT: {
+                const bool use_ckpt = common_context_can_seq_rm(ctx_dft) == COMMON_CONTEXT_SEQ_RM_TYPE_FULL;
+
                 impls.push_back(std::make_unique<common_speculative_state_draft>(config.type,
                     /* .ctx_tgt      = */ ctx_tgt,
                     /* .ctx_dft      = */ ctx_dft,
-                    /* .replacements = */ params.replacements
+                    /* .replacements = */ params.draft.replacements,
+                    /* .use_ckpt     = */ use_ckpt
                 ));
                 break;
             }
@@ -920,18 +1059,18 @@ common_speculative * common_speculative_init(
                 break;
             }
             case COMMON_SPECULATIVE_TYPE_NGRAM_SIMPLE: {
-                common_ngram_map ngram_map = get_common_ngram_map(config);
+                common_ngram_map ngram_map = get_common_ngram_map(config.type, config.params.ngram_simple);
 
                 uint16_t ngram_size_key   = ngram_map.size_key;
                 uint16_t mgram_size_value = ngram_map.size_value;
 
                 auto config_simple = common_ngram_simple_config {
-                    /* .size_ngram      = */ ngram_size_key,
-                    /* .size_mgram      = */ mgram_size_value
+                    /* .size_ngram = */ ngram_size_key,
+                    /* .size_mgram = */ mgram_size_value
                 };
                 auto state = std::make_unique<common_speculative_state_ngram_simple>(
-                    /* .type            = */ config.type,
-                    /* .state           = */ config_simple
+                    /* .type  = */ config.type,
+                    /* .state = */ config_simple
                 );
                 impls.push_back(std::move(state));
                 break;
@@ -940,18 +1079,17 @@ common_speculative * common_speculative_init(
             case COMMON_SPECULATIVE_TYPE_NGRAM_MAP_K4V: {
                 impls.push_back(std::make_unique<common_speculative_state_ngram_map_k>(
                     (config.type),
-                    get_common_ngram_map(config)
+                    get_common_ngram_map(config.type, config.params.ngram_map_k)
                 ));
                 break;
             }
             case COMMON_SPECULATIVE_TYPE_NGRAM_MOD: {
-                GGML_ASSERT(config.params.ngram_mod);
-                impls.push_back(std::make_unique<common_speculative_state_ngram_mod>(config.type, *config.params.ngram_mod));
+                GGML_ASSERT(config.params.ngram_mod.obj);
+                impls.push_back(std::make_unique<common_speculative_state_ngram_mod>(config.type, *config.params.ngram_mod.obj));
                 break;
             }
             case COMMON_SPECULATIVE_TYPE_NGRAM_CACHE: {
-                auto state = create_state_ngram_cache(
-                        params.lookup_cache_static, params.lookup_cache_dynamic, config);
+                auto state = create_state_ngram_cache(params.ngram_cache.lookup_cache_static, params.ngram_cache.lookup_cache_dynamic, config);
                 impls.push_back(std::make_unique<common_speculative_state_ngram_cache>(state));
                 break;
             }
@@ -966,7 +1104,8 @@ common_speculative * common_speculative_init(
     }
 
     auto * result = new common_speculative {
-        /* .impls = */ std::move(impls)
+        /* .impls     = */ std::move(impls),
+        /* .curr_impl = */ nullptr,
     };
 
     return result;
@@ -1008,6 +1147,15 @@ llama_tokens common_speculative_draft(
             impl->n_call_draft++;
         }
 
+        {
+            const int n_min = impl->n_min(params);
+
+            if (!result.empty() && (int) result.size() < n_min) {
+                LOG_DBG("%s: ignoring small draft: %d < %d\n", __func__, (int) result.size(), n_min);
+                result.clear();
+            }
+        }
+
         if (!result.empty()) {
             LOG_DBG("%s: called impl %s, hist size = %zu, call_count = %zu, gen = %zu\n", __func__,
                     common_speculative_type_to_str(impl.get()->type).c_str(), prompt_tgt.size(),
@@ -1017,7 +1165,7 @@ llama_tokens common_speculative_draft(
             impl->n_gen_drafts++;
             impl->n_gen_tokens += result.size();
 
-            break; // We have a draft, so break out of the loop and return it.
+            break; // we have a draft, so break out of the loop and return it.
         }
     }
 
@@ -1045,6 +1193,32 @@ void common_speculative_accept(common_speculative * spec, uint16_t n_accepted) {
     }
 }
 
+int32_t common_speculative_n_max(const common_speculative * spec, const common_params_speculative & params) {
+    if (spec == nullptr) {
+        return 0;
+    }
+
+    int32_t n_max = 0;
+    for (const auto & impl : spec->impls) {
+        n_max = std::max(n_max, impl->n_max(params));
+    }
+
+    return n_max;
+}
+
+int32_t common_speculative_n_min(const common_speculative * spec, const common_params_speculative & params) {
+    if (spec == nullptr) {
+        return 0;
+    }
+
+    int32_t n_min = 0;
+    for (const auto & impl : spec->impls) {
+        n_min = std::max(n_min, impl->n_min(params));
+    }
+
+    return n_min;
+}
+
 void common_speculative_print_stats(const common_speculative * spec) {
     if (spec == nullptr) {
         return;

common/speculative.h

@@ -14,10 +14,6 @@ enum common_speculative_type common_speculative_type_from_name(const std::string
 // convert type to string
 std::string common_speculative_type_to_str(enum common_speculative_type type);
 
-// check if the llama_context is compatible for speculative decoding
-// note: clears the memory of the context
-bool common_speculative_is_compat(llama_context * ctx_tgt);
-
 common_speculative * common_speculative_init(
         common_params_speculative & params,
         llama_context             * ctx_tgt);
@@ -37,5 +33,14 @@ llama_tokens common_speculative_draft(
 // informs the speculative decoder that n_accepted tokens were accepted by the target model
 void common_speculative_accept(common_speculative * spec, uint16_t n_accepted);
 
+int32_t common_speculative_n_max(const common_speculative * spec, const common_params_speculative & params);
+int32_t common_speculative_n_min(const common_speculative * spec, const common_params_speculative & params);
+
 // print statistics about the speculative decoding
 void common_speculative_print_stats(const common_speculative * spec);
+
+struct common_speculative_deleter {
+    void operator()(common_speculative * s) { common_speculative_free(s); }
+};
+
+typedef std::unique_ptr<common_speculative, common_speculative_deleter> common_speculative_ptr;

convert_hf_to_gguf.py

@@ -272,6 +272,22 @@ class ModelBase:
 
         return tensors
 
+    @staticmethod
+    def _scale_is_trivial(scale: Tensor) -> bool:
+        return scale.numel() <= 1 and abs(float(scale.float().sum()) - 1.0) < 1e-6
+
+    def _write_scale_tensor(self, scale_name: str, scale: Tensor):
+        if not self._scale_is_trivial(scale):
+            scale_f32 = scale.float().numpy().flatten()
+            logger.info(f"  + {scale_name} (per-tensor scale, shape [{scale_f32.size}])")
+            self.gguf_writer.add_tensor(scale_name, scale_f32)
+
+    def _write_scales_tensor(self, scale_name: str, scales: list[float]):
+        if not np.allclose(scales, 1.0, atol=1e-6):
+            scale_vals = np.array(scales, dtype=np.float32)
+            logger.info(f"  + {scale_name} (per-expert scale, shape [{len(scales)}])")
+            self.gguf_writer.add_tensor(scale_name, scale_vals)
+
     def dequant_model(self):
         # If all quantized tensors were already handled (e.g. pure NVFP4), skip
         if self._is_nvfp4 and not any(k.endswith((".weight_scale", ".weight_scale_inv")) for k in self.model_tensors):
@@ -494,7 +510,7 @@ class ModelBase:
                         s = self.model_tensors[name]
                         self.model_tensors[weight_name] = lambda w=w, s=s: dequant_simple(w(), s(), None)
                         tensors_to_remove.append(name)
-                    if name.endswith((".k_scale", ".v_scale")):
+                    if name.endswith((".input_scale", ".k_scale", ".v_scale")):
                         tensors_to_remove.append(name)
             elif quant_method is not None:
                 raise NotImplementedError(f"Quant method is not yet supported: {quant_method!r}")
@@ -602,10 +618,6 @@ class ModelBase:
         raw = np.concatenate([d_grouped, qs_grouped], axis=-1).reshape(out_features, n_super * 36)
         return raw, [out_features, n_super * 64]
 
-    @staticmethod
-    def _nvfp4_scale2_is_trivial(scale2: Tensor) -> bool:
-        return scale2.numel() <= 1 and abs(float(scale2.float().sum()) - 1.0) < 1e-6
-
     def _repack_nvfp4(self, name: str, weight: Tensor, scale: Tensor, scale2: Tensor, input_scale: Tensor):
         if "language_model." in name:
             name = name.replace("language_model.", "")
@@ -616,19 +628,8 @@ class ModelBase:
         logger.info(f"Repacked {new_name} with shape {shape} and quantization NVFP4")
         self.gguf_writer.add_tensor(new_name, raw, raw_dtype=gguf.GGMLQuantizationType.NVFP4)
 
-        # Emit per-tensor scale2 as a separate F32 tensor when non-trivial
-        if not self._nvfp4_scale2_is_trivial(scale2):
-            scale2_f32 = scale2.float().numpy().flatten()
-            scale_name = new_name.replace(".weight", ".scale")
-            logger.info(f"  + {scale_name} (per-tensor NVFP4 scale2, shape [{scale2_f32.size}])")
-            self.gguf_writer.add_tensor(scale_name, scale2_f32)
-
-        # Emit per-tensor input_scale as a separate F32 tensor when non-trivial
-        if not self._nvfp4_scale2_is_trivial(input_scale):
-            input_scale_f32 = input_scale.float().numpy().flatten()
-            input_scale_name = new_name.replace(".weight", ".input_scale")
-            logger.info(f"  + {input_scale_name} (per-tensor NVFP4 input_scale, shape [{input_scale_f32.size}])")
-            self.gguf_writer.add_tensor(input_scale_name, input_scale_f32)
+        self._write_scale_tensor(new_name.replace(".weight", ".scale"), scale2)
+        self._write_scale_tensor(new_name.replace(".weight", ".input_scale"), input_scale)
 
     def _generate_nvfp4_tensors(self):
         # Per-layer expert merging to avoid holding all experts in memory
@@ -719,24 +720,17 @@ class ModelBase:
         logger.info(f"Repacked {new_name} with shape [{len(experts)}, {shape[0]}, {shape[1]}] and quantization NVFP4")
         self.gguf_writer.add_tensor(new_name, merged, raw_dtype=gguf.GGMLQuantizationType.NVFP4)
 
-        # Emit per-expert scale2 tensor if any expert has non-trivial scale2
         scales.sort(key=lambda x: x[0])
-        scale_vals = np.array([s[1] for s in scales], dtype=np.float32)
-        if not np.allclose(scale_vals, 1.0, atol=1e-6):
-            scale_name = new_name.replace(".weight", ".scale")
-            logger.info(f"  + {scale_name} (per-expert NVFP4 scale2, shape [{len(scales)}])")
-            self.gguf_writer.add_tensor(scale_name, scale_vals)
+        self._write_scales_tensor(new_name.replace(".weight", ".scale"), [s[1] for s in scales])
 
-        # Emit per-expert input_scale tensor if any expert has non-trivial input_scale
         input_scales.sort(key=lambda x: x[0])
-        input_scale_vals = np.array([s[1] for s in input_scales], dtype=np.float32)
-        if not np.allclose(input_scale_vals, 1.0, atol=1e-6):
-            input_scale_name = new_name.replace(".weight", ".input_scale")
-            logger.info(f"  + {input_scale_name} (per-expert NVFP4 input_scale, shape [{len(input_scales)}])")
-            self.gguf_writer.add_tensor(input_scale_name, input_scale_vals)
+        self._write_scales_tensor(new_name.replace(".weight", ".input_scale"), [s[1] for s in input_scales])
 
         del experts, merged
 
+    def _needs_nvfp4_processing(self) -> bool:
+        return True
+
     def prepare_tensors(self):
         # detect NVFP4 quantization (ModelOpt format)
         quant_algo = (self.hparams.get("quantization_config") or {}).get("quant_algo")
@@ -746,7 +740,12 @@ class ModelBase:
 
         if (not quant_algo or not quant_layers) and quant_config_file.is_file():
             with open(quant_config_file, "r", encoding="utf-8") as f:
-                quant_config = json.load(f).get("quantization") or {}
+                hf_quant_config = json.load(f)
+                quant_config = hf_quant_config.get("quantization") or {}
+                producer = hf_quant_config.get("producer") or {}
+                producer_name = (producer.get("name") or "").lower()
+                if quant_method is None:
+                    self.hparams.setdefault("quantization_config", {})["quant_method"] = producer_name
                 quant_algo = quant_config.get("quant_algo", quant_algo)
                 quant_layers = quant_config.get("quantized_layers", quant_layers) or {}
 
@@ -762,7 +761,7 @@ class ModelBase:
         # NVFP4 weights are repacked and written directly to gguf_writer.
         # This must run before dequant_model so NVFP4 tensors are removed
         # from model_tensors, leaving only non-NVFP4 (e.g. FP8) for dequant.
-        if self._is_nvfp4:
+        if self._is_nvfp4 and self._needs_nvfp4_processing():
             self._generate_nvfp4_tensors()
 
         self.dequant_model()
@@ -1850,20 +1849,28 @@ class TextModel(ModelBase):
             with open(module_path, encoding="utf-8") as f:
                 modules = json.load(f)
             for mod in modules:
-                if mod["type"] == "sentence_transformers.models.Pooling":
+                if mod["type"].endswith("Pooling"):
                     pooling_path = mod["path"]
                     break
 
+        mode_mapping = {
+            "mean": gguf.PoolingType.MEAN,
+            "cls": gguf.PoolingType.CLS,
+            "lasttoken": gguf.PoolingType.LAST,
+        }
+
         # get pooling type
         if pooling_path is not None:
             with open(self.dir_model / pooling_path / "config.json", encoding="utf-8") as f:
                 pooling = json.load(f)
-            if pooling["pooling_mode_mean_tokens"]:
+            if pooling.get("pooling_mode_mean_tokens"):
                 pooling_type = gguf.PoolingType.MEAN
-            elif pooling["pooling_mode_cls_token"]:
+            elif pooling.get("pooling_mode_cls_token"):
                 pooling_type = gguf.PoolingType.CLS
-            elif pooling["pooling_mode_lasttoken"]:
+            elif pooling.get("pooling_mode_lasttoken"):
                 pooling_type = gguf.PoolingType.LAST
+            elif (pooling_mode := pooling.get("pooling_mode")) in mode_mapping:
+                pooling_type = mode_mapping[pooling_mode]
             else:
                 raise NotImplementedError("Only MEAN, CLS, and LAST pooling types supported")
             self.gguf_writer.add_pooling_type(pooling_type)
@@ -2186,6 +2193,10 @@ class MmprojModel(ModelBase):
                 # merge configs
                 self.preprocessor_config = {**self.preprocessor_config, **cfg}
 
+    def _needs_nvfp4_processing(self) -> bool:
+        # nvfp4 quantization applies to the text model only.
+        return False
+
     def get_vision_config(self) -> dict[str, Any] | None:
         config_name = "vision_config" if not self.is_mistral_format else "vision_encoder"
         return self.global_config.get(config_name)
@@ -4446,6 +4457,12 @@ class NemotronNanoV2VLModel(MmprojModel):
         }
         return vision_config
 
+    def dequant_model(self):
+        if self._is_nvfp4:
+            # Skip nvfp4 quantization for vision/audio model.
+            return
+        super().dequant_model()
+
     def set_gguf_parameters(self):
         if "image_mean" not in self.preprocessor_config:
             self.preprocessor_config["image_mean"] = [0.485, 0.456, 0.406]
@@ -4469,6 +4486,10 @@ class NemotronNanoV2VLModel(MmprojModel):
         if "input_conditioner" in name:
             return
 
+        # mtmd does not support video yet so skip tensors related to video.
+        if "radio_model.model.patch_generator.video_embedder" in name:
+            return
+
         # RADIO's pos_embed doesn't have .weight suffix, but clip.cpp expects it
         if "patch_generator.pos_embed" in name:
             if not name.endswith(".weight"):
@@ -6637,7 +6658,7 @@ class BertModel(TextModel):
 
         tokens: list[bytes] = [f"[PAD{i}]".encode("utf-8") for i in range(vocab_size)]
         scores: list[float] = [-10000.0] * vocab_size
-        toktypes: list[int] = [SentencePieceTokenTypes.UNUSED] * vocab_size  # ty: ignore[invalid-assignment]
+        toktypes: list[int] = [SentencePieceTokenTypes.UNUSED] * vocab_size
 
         if isinstance(tokenizer, SentencePieceProcessor):
             for token_id in range(tokenizer.vocab_size()):
@@ -7180,7 +7201,7 @@ class EmbeddingGemma(Gemma3Model):
                 with open(modules_file, encoding="utf-8") as modules_json_file:
                     mods = json.load(modules_json_file)
                 for mod in mods:
-                    if mod["type"] == "sentence_transformers.models.Dense":
+                    if mod["type"].endswith("Dense"):
                         mod_path = mod["path"]
                         # check if model.safetensors file for Dense layer exists
                         model_tensors_file = self.dir_model / mod_path / "model.safetensors"
@@ -10816,7 +10837,11 @@ class NemotronHModel(GraniteHybridModel):
         # uses self.model_arch to build the tensor name map, and all MoE-specific
         # mappings would be missed if it were called with the default non-MoE arch.
         hparams = ModelBase.load_hparams(args[0], self.is_mistral_format)
-        if "num_experts_per_tok" in hparams:
+        has_moe_params = (
+            "num_experts_per_tok" in hparams
+            or (isinstance(hparams.get("llm_config"), dict) and "num_experts_per_tok" in hparams["llm_config"])
+        )
+        if has_moe_params:
             self.model_arch = gguf.MODEL_ARCH.NEMOTRON_H_MOE
             self.is_moe = True
 
@@ -10893,7 +10918,64 @@ class NemotronHModel(GraniteHybridModel):
                 self.gguf_writer.add_moe_latent_size(latent_size)
 
     def set_vocab(self):
-        super().set_vocab()
+        # The NemotronH config uses pattern characters (e.g. '-') that may not
+        # be supported by the installed transformers version. AutoTokenizer
+        # internally calls AutoConfig which triggers this parsing failure.
+        # Using trust_remote_code=True to load the model's own config class.
+        tokens: list[str] = []
+        toktypes: list[int] = []
+
+        from transformers import AutoTokenizer
+        tokenizer = AutoTokenizer.from_pretrained(self.dir_model, trust_remote_code=True)
+
+        # Pad vocab size (from Mamba2Model/GraniteHybridModel)
+        self.hparams["pad_vocab_size_multiple"] = 8 # Setting this here since GraniteHybridModel.set_vocab() isn't being invoked now.
+        # From Mamba2Model.set_vocab():
+        vocab_size = self.hparams["vocab_size"]
+        pad_vocab = self.hparams.get("pad_vocab_size_multiple", 16)
+        # ref: https://stackoverflow.com/a/17511341/22827863
+        vocab_size = -(vocab_size // -pad_vocab) * pad_vocab
+        self.hparams["vocab_size"] = vocab_size
+
+        assert max(tokenizer.vocab.values()) < vocab_size  # ty: ignore[unresolved-attribute]
+
+        tokpre = self.get_vocab_base_pre(tokenizer)
+
+        reverse_vocab = {id_: encoded_tok for encoded_tok, id_ in tokenizer.vocab.items()}  # ty: ignore[unresolved-attribute]
+        added_vocab = tokenizer.get_added_vocab()  # ty: ignore[unresolved-attribute]
+
+        added_tokens_decoder = tokenizer.added_tokens_decoder  # ty: ignore[unresolved-attribute]
+
+        for i in range(vocab_size):
+            if i not in reverse_vocab:
+                tokens.append(f"[PAD{i}]")
+                toktypes.append(gguf.TokenType.UNUSED)
+            else:
+                token: str = reverse_vocab[i]
+                if token in added_vocab:
+                    if not added_tokens_decoder[i].normalized:
+                        previous_token = token
+                        token = tokenizer.decode(tokenizer.encode(token, add_special_tokens=False))  # ty: ignore[unresolved-attribute, invalid-assignment]
+                        if previous_token != token:
+                            logger.info(f"{repr(previous_token)} is encoded and decoded back to {repr(token)} using AutoTokenizer")
+
+                    if added_tokens_decoder[i].special or self.does_token_look_special(token):
+                        toktypes.append(gguf.TokenType.CONTROL)
+                    else:
+                        token = token.replace(b"\xe2\x96\x81".decode("utf-8"), " ")  # pre-normalize user-defined spaces
+                        toktypes.append(gguf.TokenType.USER_DEFINED)
+                else:
+                    toktypes.append(gguf.TokenType.NORMAL)
+                tokens.append(token)
+
+        # From TextModel.set_vocab_gpt2():
+        self.gguf_writer.add_tokenizer_model("gpt2")
+        self.gguf_writer.add_tokenizer_pre(tokpre)
+        self.gguf_writer.add_token_list(tokens)
+        self.gguf_writer.add_token_types(toktypes)
+
+        special_vocab = gguf.SpecialVocab(self.dir_model, load_merges=True)
+        special_vocab.add_to_gguf(self.gguf_writer)
 
         # The tokenizer _does_ add a BOS token (via post_processor type
         # TemplateProcessing) but does not set add_bos_token to true in the
@@ -10906,6 +10988,11 @@ class NemotronHModel(GraniteHybridModel):
         if name.startswith(("vision_model.", "mlp1.")):
             return
 
+        if name.startswith(("sound_encoder.")):
+            return
+        if name.startswith(("sound_projection.")):
+            return
+
         # Strip language_model. prefix for VLM models (e.g., Nemotron Nano 12B v2 VL)
         if name.startswith("language_model."):
             name = name[len("language_model."):]
@@ -11790,7 +11877,7 @@ class LLaDAMoEModel(TextModel):
                 raise ValueError(f"Unprocessed experts: {experts}")
 
 
-@ModelBase.register("HunYuanDenseV1ForCausalLM", "HunYuanVLForConditionalGeneration")
+@ModelBase.register("HunYuanDenseV1ForCausalLM")
 class HunYuanModel(TextModel):
     model_arch = gguf.MODEL_ARCH.HUNYUAN_DENSE
 
@@ -11929,28 +12016,58 @@ class HunYuanModel(TextModel):
 
 
 @ModelBase.register("HunYuanVLForConditionalGeneration")
-class HunyuanOCRVisionModel(MmprojModel):
+class HunyuanVLVisionModel(MmprojModel):
+    # Handles both HunyuanOCR and HunyuanVL, which share the HF architecture name
+    # "HunYuanVLForConditionalGeneration" and the `vit.perceive.*` vision layout.
+    # Each variant maps to a different projector type in clip.cpp so image
+    # preprocessing follows the correct code path.
+
     def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)
         assert self.hparams_vision is not None
-        # HunyuanOCR uses max_image_size instead of image_size
+        # HunyuanOCR / HunyuanVL uses max_image_size instead of image_size
         if "image_size" not in self.hparams_vision:
             self.hparams_vision["image_size"] = self.hparams_vision.get("max_image_size", 2048)
 
+    @staticmethod
+    def is_ocr_variant(hparams: dict) -> bool:
+        """Return True for HunyuanOCR, False for HunyuanVL.
+
+        The projector's output dim must equal the text model's hidden_size by
+        construction (that's what "projector" means). HunyuanOCR pairs a 1B text
+        backbone (hidden=1024); HunyuanVL pairs a 4B one (hidden=3072). So the
+        ViT -> LLM projection dim is a hard architectural signature, not a
+        magic number.
+        """
+        vision_out = int((hparams.get("vision_config") or {}).get("out_hidden_size", 0))
+        return vision_out == 1024
+
     def set_gguf_parameters(self):
         super().set_gguf_parameters()
         assert self.hparams_vision is not None
-        hparams = self.hparams_vision
-        self.gguf_writer.add_clip_projector_type(gguf.VisionProjectorType.HUNYUANOCR)
-        self.gguf_writer.add_vision_use_gelu(True)
-        self.gguf_writer.add_vision_attention_layernorm_eps(hparams.get("rms_norm_eps", 1e-5))
-        self.gguf_writer.add_vision_spatial_merge_size(hparams.get("spatial_merge_size", 2))
-        self.gguf_writer.add_vision_min_pixels(self.preprocessor_config["min_pixels"])
-        self.gguf_writer.add_vision_max_pixels(self.preprocessor_config["max_pixels"])
+        vcfg = self.hparams_vision
+
+        if self.is_ocr_variant(self.global_config):
+            # --- HunyuanOCR ---
+            self.gguf_writer.add_clip_projector_type(gguf.VisionProjectorType.HUNYUANOCR)
+            self.gguf_writer.add_vision_use_gelu(True)
+            self.gguf_writer.add_vision_attention_layernorm_eps(vcfg.get("rms_norm_eps", 1e-5))
+            self.gguf_writer.add_vision_spatial_merge_size(vcfg.get("spatial_merge_size", 2))
+            self.gguf_writer.add_vision_min_pixels(self.preprocessor_config["min_pixels"])
+            self.gguf_writer.add_vision_max_pixels(self.preprocessor_config["max_pixels"])
+            return
+
+        # --- HunyuanVL ---
+        self.gguf_writer.add_clip_projector_type(gguf.VisionProjectorType.HUNYUANVL)
+        self.gguf_writer.add_vision_use_gelu(str(vcfg["hidden_act"]).lower() == "gelu")
+        self.gguf_writer.add_vision_attention_layernorm_eps(float(vcfg["rms_norm_eps"]))
+        self.gguf_writer.add_vision_spatial_merge_size(int(vcfg["spatial_merge_size"]))
+        self.gguf_writer.add_vision_min_pixels(int(self.preprocessor_config["min_pixels"]))
+        self.gguf_writer.add_vision_max_pixels(int(self.preprocessor_config["max_pixels"]))
 
     def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
         if not name.startswith("vit."):
-            return  # skip text tensors
+            return
         # strip CLS token (row 0) from position embeddings so resize_position_embeddings works
         if "position_embedding" in name:
             data_torch = data_torch[1:]  # [n_patches+1, n_embd] -> [n_patches, n_embd]
@@ -11958,11 +12075,66 @@ class HunyuanOCRVisionModel(MmprojModel):
 
     def tensor_force_quant(self, name, new_name, bid, n_dims):
         # force conv weights to F32 or F16 to avoid BF16 IM2COL issues on Metal
+        # Both HunyuanOCR and HunyuanVL emit the ViT -> LLM projection as mm.0/mm.2.
         if ("mm.0." in new_name or "mm.2." in new_name) and new_name.endswith(".weight"):
             return gguf.GGMLQuantizationType.F16 if self.ftype == gguf.LlamaFileType.MOSTLY_F16 else gguf.GGMLQuantizationType.F32
         return super().tensor_force_quant(name, new_name, bid, n_dims)
 
 
+@ModelBase.register("HunYuanVLForConditionalGeneration")
+class HunyuanVLTextModel(HunYuanModel):
+    # The "HunYuanVLForConditionalGeneration" HF architecture covers both HunyuanOCR
+    # and HunyuanVL. HunyuanOCR reuses the HunYuan-Dense text backbone (standard RoPE),
+    # while HunyuanVL introduces a new LLM arch with XD-RoPE. Detect the variant from
+    # the config and pick the matching GGUF architecture.
+    model_arch = gguf.MODEL_ARCH.HUNYUAN_VL
+
+    @staticmethod
+    def _is_ocr_config(hparams: dict) -> bool:
+        # OCR pairs a 1B text backbone (hidden=1024) with a ViT projector that
+        # outputs 1024-d; HunyuanVL uses 3072-d. Keep in sync with
+        # HunyuanVLVisionModel.is_ocr_variant.
+        return int((hparams.get("vision_config") or {}).get("out_hidden_size", 0)) == 1024
+
+    def __init__(self, dir_model: Path, *args, **kwargs):
+        raw_hparams = kwargs.get("hparams") or ModelBase.load_hparams(dir_model, is_mistral_format=False)
+        if self._is_ocr_config(raw_hparams):
+            self.model_arch = gguf.MODEL_ARCH.HUNYUAN_DENSE
+        else:
+            self.model_arch = gguf.MODEL_ARCH.HUNYUAN_VL
+        super().__init__(dir_model, *args, **kwargs)
+
+    def set_gguf_parameters(self):
+        super().set_gguf_parameters()
+
+        # Only emit XD-RoPE metadata for the HunyuanVL backbone; HunyuanOCR uses
+        # the HunYuan-Dense arch which already handles standard rope in super().
+        if self.model_arch != gguf.MODEL_ARCH.HUNYUAN_VL:
+            return
+
+        if self.rope_parameters.get("rope_type") != "xdrope":
+            return
+
+        # defaults for HunyuanVL. The C++ side later computes:
+        #   freq_base = rope_theta * alpha ** (head_dim / (head_dim - 2))
+        self.gguf_writer.add_rope_freq_base(float(self.rope_parameters["rope_theta"]))
+        self.gguf_writer.add_rope_scaling_alpha(float(self.rope_parameters["alpha"]))
+        self.gguf_writer.add_rope_scaling_type(gguf.RopeScalingType.NONE)
+        self.gguf_writer.add_rope_scaling_factor(float(self.rope_parameters.get("factor", 1)))
+
+        ctx_len = int(self.hparams["max_position_embeddings"])
+        self.gguf_writer.add_rope_scaling_orig_ctx_len(ctx_len)
+        self.gguf_writer.add_context_length(ctx_len)
+
+        self.gguf_writer.add_rope_dimension_sections(list(self.rope_parameters["xdrope_section"]))
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        # Skip vision tensors — they are written by HunyuanVLVisionModel
+        if name.startswith("vit."):
+            return
+        yield from super().modify_tensors(data_torch, name, bid)
+
+
 @ModelBase.register("SmolLM3ForCausalLM")
 class SmolLM3Model(LlamaModel):
     model_arch = gguf.MODEL_ARCH.SMOLLM3
@@ -13060,17 +13232,18 @@ class LazyTorchTensor(gguf.LazyBase):
     }
 
     # only used when byteswapping data. Only correct size is needed
+    # TODO: uncomment uint64, uint32, and uint16, ref: https://github.com/pytorch/pytorch/issues/58734
     _dtype_byteswap_map: dict[torch.dtype, type] = {
         torch.float64: np.float64,
         torch.float32: np.float32,
         torch.bfloat16: np.float16,
         torch.float16: np.float16,
         torch.int64: np.int64,
-        torch.uint64: np.uint64,
+        # torch.uint64: np.uint64,
         torch.int32: np.int32,
-        torch.uint32: np.uint32,
+        # torch.uint32: np.uint32,
         torch.int16: np.int16,
-        torch.uint16: np.uint16,
+        # torch.uint16: np.uint16,
         torch.int8: np.int8,
         torch.uint8: np.uint8,
         torch.bool: np.uint8,

docs/backend/OPENVINO.md

@@ -244,7 +244,6 @@ build\ReleaseOV\bin\llama-cli.exe -m "C:\models\Llama-3.2-1B-Instruct-Q4_0.gguf"
 - `-fa 1` is required when running llama-bench with the OpenVINO backend.
   - `GGML_OPENVINO_STATEFUL_EXECUTION=1 GGML_OPENVINO_DEVICE=GPU ./llama-bench -fa 1`
 - `llama-server` with OpenVINO backend supports only one chat session/thread, when `GGML_OPENVINO_STATEFUL_EXECUTION=1` is enabled.
-- For Intel GPU, NPU detection in containers, GPU, NPU user-space drivers/libraries must be present inside the image. We will include in a future PR. Until then, you can use this reference Dockerfile: [openvino.Dockerfile](https://github.com/ravi9/llama.cpp/blob/ov-docker-update/.devops/openvino.Dockerfile)
 
 > [!NOTE]
 > The OpenVINO backend is actively under development. Fixes are underway, and this document will continue to be updated as issues are resolved.
@@ -274,8 +273,6 @@ docker build --build-arg http_proxy=$http_proxy --build-arg https_proxy=$https_p
 Run llama.cpp with OpenVINO backend Docker container.
 Save sample models in `~/models` as [shown above](#3-download-sample-model). It will be mounted to the container in the examples below.
 
-> [!NOTE]
-> Intel GPU, NPU detection in containers will be included in a future PR. Until then, you can use this reference Dockerfile: [openvino.Dockerfile](https://github.com/ravi9/llama.cpp/blob/ov-docker-update/.devops/openvino.Dockerfile).
 
 ```bash
 #  Run Docker container

docs/backend/SYCL.md

@@ -31,6 +31,8 @@ SYCL cross-platform capabilities enable support for other vendor GPUs as well.
 
 ## Recommended Release
 
+### Windows
+
 The following releases are verified and recommended:
 
 |Commit ID|Tag|Release|Verified  Platform| Update date|
@@ -39,9 +41,22 @@ The following releases are verified and recommended:
 |3bcd40b3c593d14261fb2abfabad3c0fb5b9e318|b4040 |[llama-b4040-bin-win-sycl-x64.zip](https://github.com/ggml-org/llama.cpp/releases/download/b4040/llama-b4040-bin-win-sycl-x64.zip) |Arc A770/Linux/oneAPI 2024.1<br>MTL Arc GPU/Windows 11/oneAPI 2024.1| 2024-11-19|
 |fb76ec31a9914b7761c1727303ab30380fd4f05c|b3038 |[llama-b3038-bin-win-sycl-x64.zip](https://github.com/ggml-org/llama.cpp/releases/download/b3038/llama-b3038-bin-win-sycl-x64.zip) |Arc A770/Linux/oneAPI 2024.1<br>MTL Arc GPU/Windows 11/oneAPI 2024.1||
 
+### Ubuntu 24.04
+
+The release packages for Ubuntu 24.04 x64 (FP32/FP16) only include the binary files of the llama.cpp SYCL backend. They require the target machine to have pre-installed Intel GPU drivers and oneAPI packages that are the same version as the build package. To get the version and installation info, refer to release.yml: ubuntu-24-sycl -> Download & Install oneAPI.
+
+It is recommended to use them with Intel Docker.
+
+The packages for FP32 and FP16 would have different accuracy and performance on LLMs. Please choose it acording to the test result.
 
 ## News
 
+- 2026.04
+
+  - Optimize mul_mat by reorder feature for data type: Q4_K, Q5_K, Q_K, Q8_0.
+  - Fused MoE.
+  - Upgrate CI and built package for oneAPI 2025.3.3, support Ubuntu 24.04 built package.
+
 - 2026.03
   - Support Flash-Attention: less memory usage, performance impact depends on LLM.
 
@@ -229,6 +244,7 @@ Upon a successful installation, SYCL is enabled for the available intel devices,
 
 |Verified release|
 |-|
+|2025.3.3 |
 |2025.2.1|
 |2025.1|
 |2024.1|
@@ -339,6 +355,12 @@ Choose one of following methods to run.
 ./examples/sycl/test.sh
 ```
 
+- Run llama-server:
+
+```sh
+./examples/sycl/start-svr.sh -m PATH/MODEL_FILE
+```
+
 2. Command line
 Launch inference
 
@@ -627,10 +649,18 @@ Choose one of following methods to run.
 
 1. Script
 
+- Run test:
+
 ```
 examples\sycl\win-test.bat
 ```
 
+- Run llama-server:
+
+```
+examples\sycl\win-start-svr.bat -m PATH\MODEL_FILE
+```
+
 2. Command line
 
 Launch inference
@@ -689,6 +719,7 @@ use 1 SYCL GPUs: [0] with Max compute units:512
 | GGML_SYCL_F16      | OFF *(default)* \|ON *(optional)*     | Enable FP16 build with SYCL code path. (1.) |
 | GGML_SYCL_GRAPH    | OFF *(default)* \|ON *(Optional)*     | Enable build with [SYCL Graph extension](https://github.com/intel/llvm/blob/sycl/sycl/doc/extensions/experimental/sycl_ext_oneapi_graph.asciidoc). |
 | GGML_SYCL_DNN      | ON *(default)* \|OFF *(Optional)*     | Enable build with oneDNN.                   |
+| GGML_SYCL_HOST_MEM_FALLBACK | ON *(default)* \|OFF *(Optional)* | Allow host memory fallback when device memory is full during quantized weight reorder. Enables inference to continue at reduced speed (reading over PCIe) instead of failing. Requires Linux kernel 6.8+. |
 | CMAKE_C_COMPILER   | `icx` *(Linux)*, `icx/cl` *(Windows)* | Set `icx` compiler for SYCL code path.      |
 | CMAKE_CXX_COMPILER | `icpx` *(Linux)*, `icx` *(Windows)*   | Set `icpx/icx` compiler for SYCL code path. |
 

docs/backend/snapdragon/README.md

@@ -249,18 +249,27 @@ build: 6a8cf8914 (6733)
   ```
 
 - `GGML_HEXAGON_PROFILE=1`
-  Generates a host-side profile for the ggml-hexagon Ops.
+  Enables Op profiling:
 
-- `GGML_HEXAGON_OPMASK=0x0`
-  Allows enabling specific stages of the processing pipeline:
+  - `1` Basic profile with per-op `usecs` and `cycles` counters
+  - `2` Extended profile with per-op `usecs`, `cycles` and default PMU counter data
+  - `0x1,...,0x8` Extended profile with per-op `usecs`, `cycles` and custom PMU counter data
+
+  The logging output can be either saved into a file for post-processing or it can be piped directly into the post-processing tool to generate the report.
+  Examples:
+
+      `GGML_HEXAGON_PROFILE=1 llama-completion ... |& ./scripts/snapdragon/ggml-hexagon-profile.py -`
+
+- `GGML_HEXAGON_OPSTAGE=0x0`
+  Allows enabling specific stages of the Op processing pipeline:
 
   - `0x1` Enable Op Queue (i.e., queuing Ops into NPU)
   - `0x2` Enable Op Compute (MUL_MAT, etc.)
 
   Examples:
 
-      `GGML_HEXAGON_OPMASK=0x1 llama-completion ...` - Ops are enqueued but NPU-side processing is stubbed out
-      `GGML_HEXAGON_OPMASK=0x3 llama-completion ...` - Full queuing and processing of Ops (default)
+      `GGML_HEXAGON_OPSTAGE=0x1 llama-completion ...` - Ops are enqueued to the NPU but dma & compute are disabled
+      `GGML_HEXAGON_OPSTAGE=0x3 llama-completion ...` - Full queuing and processing of Ops (default)
 
 - `GGML_HEXAGON_OPFILTER=regex`
   Allows filtering (disabling) Ops that match the regex pattern:

docs/ops.md

@@ -22,13 +22,13 @@ Legend:
 |                           ARANGE | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                           ARGMAX | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ |
 |                          ARGSORT | ❌ | ✅ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ | ❌ |
-|                             CEIL | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
+|                             CEIL | ❌ | ❌ | ✅ | 🟡 | ✅ | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
 |                            CLAMP | ❌ | ✅ | ✅ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | ❌ | ❌ |
 |                           CONCAT | ❌ | ✅ | ✅ | 🟡 | ✅ | 🟡 | ✅ | ✅ | ✅ | ❌ | ❌ |
-|                             CONT | ❌ | 🟡 | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ | ❌ |
-|                          CONV_2D | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ❌ | ❌ | ❌ |
+|                             CONT | ❌ | 🟡 | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | 🟡 | ❌ | ❌ |
+|                          CONV_2D | ❌ | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ |
 |                       CONV_2D_DW | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
-|                          CONV_3D | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                          CONV_3D | ❌ | ❌ | ✅ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
 |                CONV_TRANSPOSE_1D | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                CONV_TRANSPOSE_2D | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
 |                              COS | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | 🟡 | 🟡 | ✅ | ❌ | ❌ |
@@ -46,7 +46,7 @@ Legend:
 |                            EXPM1 | ❌ | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ |
 |                             FILL | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ |
 |                   FLASH_ATTN_EXT | ❌ | 🟡 | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ | ❌ |
-|                            FLOOR | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ✅ | ❌ | ❌ |
+|                            FLOOR | ❌ | ❌ | ✅ | 🟡 | ✅ | ❌ | 🟡 | 🟡 | ✅ | ❌ | ❌ |
 |                  GATED_DELTA_NET | ❌ | ❌ | ✅ | ❌ | 🟡 | ❌ | ✅ | ❌ | ✅ | ❌ | ❌ |
 |                GATED_LINEAR_ATTN | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ |
 |                            GEGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ | ❌ |
@@ -60,7 +60,7 @@ Legend:
 |                       GROUP_NORM | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                      HARDSIGMOID | ❌ | ✅ | ✅ | 🟡 | ✅ | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
 |                        HARDSWISH | ❌ | ✅ | ✅ | 🟡 | ✅ | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
-|                           IM2COL | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
+|                           IM2COL | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
 |                        IM2COL_3D | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ |
 |                          L2_NORM | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ |
 |                       LEAKY_RELU | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ |
@@ -84,10 +84,10 @@ Legend:
 |                      REPEAT_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                         RMS_NORM | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
 |                    RMS_NORM_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
-|                             ROLL | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
+|                             ROLL | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                             ROPE | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
 |                        ROPE_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
-|                            ROUND | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ✅ | ❌ | ❌ |
+|                            ROUND | ❌ | ❌ | ✅ | 🟡 | ✅ | ❌ | 🟡 | 🟡 | ✅ | ❌ | ❌ |
 |                        RWKV_WKV6 | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                        RWKV_WKV7 | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                            SCALE | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
@@ -105,7 +105,7 @@ Legend:
 |                              SQR | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ❌ | ❌ |
 |                             SQRT | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ❌ | ❌ |
 |                         SSM_CONV | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
-|                         SSM_SCAN | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | 🟡 | ❌ | ❌ | ❌ |
+|                         SSM_SCAN | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | 🟡 | ✅ | ❌ | ❌ |
 |                             STEP | ❌ | ✅ | ✅ | 🟡 | ✅ | ❌ | ✅ | 🟡 | ✅ | ❌ | ❌ |
 |                              SUB | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ |
 |                              SUM | ❌ | 🟡 | ✅ | 🟡 | 🟡 | ❌ | 🟡 | 🟡 | 🟡 | ❌ | ❌ |
@@ -116,6 +116,6 @@ Legend:
 |               TIMESTEP_EMBEDDING | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ |
 |                            TOP_K | ❌ | ❌ | ✅ | ❌ | ✅ | ❌ | 🟡 | 🟡 | ✅ | ❌ | ❌ |
 |                              TRI | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ |
-|                            TRUNC | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ✅ | ❌ | ❌ |
+|                            TRUNC | ❌ | ❌ | ✅ | 🟡 | ✅ | ❌ | 🟡 | 🟡 | ✅ | ❌ | ❌ |
 |                          UPSCALE | ❌ | 🟡 | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ❌ | ❌ | ❌ |
-|                            XIELU | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ |
+|                            XIELU | ❌ | ❌ | ✅ | ❌ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ |

examples/batched/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-batched)
 add_executable(${TARGET} batched.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/convert-llama2c-to-ggml/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-convert-llama2c-to-ggml)
 add_executable(${TARGET} convert-llama2c-to-ggml.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/debug/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-debug)
 add_executable(${TARGET} debug.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/debug/debug.cpp

@@ -202,10 +202,14 @@ static bool run(llama_context * ctx, const common_params & params) {
     print_tokenized_prompt(ctx, tokens, params.prompt);
 
     if (params.save_logits) {
-        output_data output {ctx, model, params};
-        std::filesystem::path model_path{params.model.path};
-        std::string model_name{model_path.stem().string()};
-        save_output_data(output, model_name, params.logits_output_dir);
+        try {
+            output_data output {ctx, model, params};
+            std::filesystem::path model_path{params.model.path};
+            std::string model_name{model_path.stem().string()};
+            save_output_data(output, model_name, params.logits_output_dir);
+        } catch (const std::exception & e) {
+            LOG_ERR("%s : error saving logits: %s\n", __func__, e.what());
+        }
     }
 
     return true;
@@ -223,7 +227,7 @@ int main(int argc, char ** argv) {
     llama_backend_init();
     llama_numa_init(params.numa);
 
-    std::optional<base_callback_data> cb_data;
+    std::optional<common_debug_cb_user_data> cb_data;
     if (!params.save_logits) {
         cb_data.emplace(params, params.tensor_filter);
     }

examples/diffusion/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-diffusion-cli)
 add_executable(${TARGET} diffusion-cli.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE llama common ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama llama-common ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/embedding/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-embedding)
 add_executable(${TARGET} embedding.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/eval-callback/CMakeLists.txt

@@ -1,7 +1,7 @@
 set(TARGET llama-eval-callback)
 add_executable(${TARGET} eval-callback.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)
 
 if(LLAMA_BUILD_TESTS)

examples/eval-callback/eval-callback.cpp

@@ -3,7 +3,6 @@
 #include "debug.h"
 #include "log.h"
 #include "llama.h"
-#include "llama-cpp.h"
 
 #include <clocale>
 #include <string>
@@ -38,7 +37,7 @@ static bool run(llama_context * ctx, const common_params & params) {
 int main(int argc, char ** argv) {
     std::setlocale(LC_NUMERIC, "C");
 
-    base_callback_data cb_data;
+    common_debug_cb_user_data cb_data;
 
     common_params params;
 
@@ -53,7 +52,7 @@ int main(int argc, char ** argv) {
 
     // pass the callback to the backend scheduler
     // it will be executed for each node during the graph computation
-    params.cb_eval = common_debug_cb_eval<false>;
+    params.cb_eval = common_debug_cb_eval;
     params.cb_eval_user_data = &cb_data;
     params.warmup = false;
 

examples/gen-docs/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-gen-docs)
 add_executable(${TARGET} gen-docs.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/gen-docs/gen-docs.cpp

@@ -73,12 +73,12 @@ static void write_help(std::ostringstream & ss, const md_file & md) {
     auto ctx_arg = common_params_parser_init(params, md.ex);
 
     std::vector<common_arg *> common_options;
-    std::vector<common_arg *> sparam_options;
+    std::vector<common_arg *> sampling_options;
     std::vector<common_arg *> specific_options;
     for (auto & opt : ctx_arg.options) {
         // in case multiple LLAMA_EXAMPLE_* are set, we prioritize the LLAMA_EXAMPLE_* matching current example
-        if (opt.is_sparam) {
-            sparam_options.push_back(&opt);
+        if (opt.is_sampling) {
+            sampling_options.push_back(&opt);
         } else if (opt.in_example(ctx_arg.ex)) {
             specific_options.push_back(&opt);
         } else {
@@ -93,7 +93,7 @@ static void write_help(std::ostringstream & ss, const md_file & md) {
     ss << "### Common params\n\n";
     write_table(ss, common_options);
     ss << "\n\n### Sampling params\n\n";
-    write_table(ss, sparam_options);
+    write_table(ss, sampling_options);
     ss << "\n\n### " << md.specific_section_header << "\n\n";
     write_table(ss, specific_options);
 

examples/idle/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-idle)
 add_executable(${TARGET} idle.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE llama common ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama llama-common ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_11)

examples/llama.android/lib/src/main/cpp/CMakeLists.txt

@@ -51,6 +51,6 @@ target_include_directories(${CMAKE_PROJECT_NAME} PRIVATE
 
 target_link_libraries(${CMAKE_PROJECT_NAME}
         llama
-        common
+        llama-common
         android
         log)

examples/lookahead/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-lookahead)
 add_executable(${TARGET} lookahead.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/lookup/CMakeLists.txt

@@ -1,23 +1,23 @@
 set(TARGET llama-lookup)
 add_executable(${TARGET} lookup.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)
 
 set(TARGET llama-lookup-create)
 add_executable(${TARGET} lookup-create.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)
 
 set(TARGET llama-lookup-merge)
 add_executable(${TARGET} lookup-merge.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)
 
 set(TARGET llama-lookup-stats)
 add_executable(${TARGET} lookup-stats.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/lookup/lookup-create.cpp

@@ -37,9 +37,9 @@ int main(int argc, char ** argv){
 
     common_ngram_cache ngram_cache;
     common_ngram_cache_update(ngram_cache, LLAMA_NGRAM_STATIC, LLAMA_NGRAM_STATIC, inp, inp.size(), true);
-    fprintf(stderr, "%s: hashing done, writing file to %s\n", __func__, params.speculative.lookup_cache_static.c_str());
+    fprintf(stderr, "%s: hashing done, writing file to %s\n", __func__, params.speculative.ngram_cache.lookup_cache_static.c_str());
 
-    common_ngram_cache_save(ngram_cache, params.speculative.lookup_cache_static);
+    common_ngram_cache_save(ngram_cache, params.speculative.ngram_cache.lookup_cache_static);
 
     return 0;
 }

examples/lookup/lookup-stats.cpp

@@ -24,7 +24,7 @@ int main(int argc, char ** argv){
         return 1;
     }
 
-    const int n_draft = params.speculative.n_max;
+    const int n_draft = params.speculative.draft.n_max;
 
     // init llama.cpp
     llama_backend_init();
@@ -49,18 +49,18 @@ int main(int argc, char ** argv){
     {
         const int64_t t_start_draft_us = ggml_time_us();
 
-        if (!params.speculative.lookup_cache_static.empty()) {
+        if (!params.speculative.ngram_cache.lookup_cache_static.empty()) {
             try {
-                ngram_cache_static = common_ngram_cache_load(params.speculative.lookup_cache_static);
+                ngram_cache_static = common_ngram_cache_load(params.speculative.ngram_cache.lookup_cache_static);
             } catch (std::ifstream::failure const &) {
-                LOG_ERR("failed to open static lookup cache: %s", params.speculative.lookup_cache_static.c_str());
+                LOG_ERR("failed to open static lookup cache: %s", params.speculative.ngram_cache.lookup_cache_static.c_str());
                 exit(1);
             }
         }
 
-        if (!params.speculative.lookup_cache_dynamic.empty()) {
+        if (!params.speculative.ngram_cache.lookup_cache_dynamic.empty()) {
             try {
-                ngram_cache_dynamic = common_ngram_cache_load(params.speculative.lookup_cache_dynamic);
+                ngram_cache_dynamic = common_ngram_cache_load(params.speculative.ngram_cache.lookup_cache_dynamic);
             } catch (std::ifstream::failure const &) {} // if the file does not exist it will simply be created at the end of the program
         }
 

examples/lookup/lookup.cpp

@@ -25,7 +25,7 @@ int main(int argc, char ** argv){
     }
 
     // max. number of additional tokens to draft if match is found
-    const int n_draft = params.speculative.n_max;
+    const int n_draft = params.speculative.draft.n_max;
 
     // init llama.cpp
     llama_backend_init();
@@ -54,18 +54,18 @@ int main(int argc, char ** argv){
         const int64_t t_start_draft_us = ggml_time_us();
         common_ngram_cache_update(ngram_cache_context, LLAMA_NGRAM_MIN, LLAMA_NGRAM_MAX, inp, inp.size(), false);
 
-        if (!params.speculative.lookup_cache_static.empty()) {
+        if (!params.speculative.ngram_cache.lookup_cache_static.empty()) {
             try {
-                ngram_cache_static = common_ngram_cache_load(params.speculative.lookup_cache_static);
+                ngram_cache_static = common_ngram_cache_load(params.speculative.ngram_cache.lookup_cache_static);
             } catch (std::ifstream::failure const &) {
-                LOG_ERR("failed to open static lookup cache: %s", params.speculative.lookup_cache_static.c_str());
+                LOG_ERR("failed to open static lookup cache: %s", params.speculative.ngram_cache.lookup_cache_static.c_str());
                 exit(1);
             }
         }
 
-        if (!params.speculative.lookup_cache_dynamic.empty()) {
+        if (!params.speculative.ngram_cache.lookup_cache_dynamic.empty()) {
             try {
-                ngram_cache_dynamic = common_ngram_cache_load(params.speculative.lookup_cache_dynamic);
+                ngram_cache_dynamic = common_ngram_cache_load(params.speculative.ngram_cache.lookup_cache_dynamic);
             } catch (std::ifstream::failure const &) {} // if the file does not exist it will simply be created at the end of the program
         }
 
@@ -213,7 +213,7 @@ int main(int argc, char ** argv){
 
     // Update dynamic ngram cache with context ngram cache and save it to disk:
     common_ngram_cache_merge(ngram_cache_dynamic, ngram_cache_context);
-    common_ngram_cache_save(ngram_cache_dynamic, params.speculative.lookup_cache_dynamic);
+    common_ngram_cache_save(ngram_cache_dynamic, params.speculative.ngram_cache.lookup_cache_dynamic);
 
     LOG("\n\n");
 

examples/model-conversion/scripts/causal/convert-model.sh

@@ -25,7 +25,11 @@ MODEL_NAME="${MODEL_NAME:-$(basename "$MODEL_PATH")}"
 OUTPUT_DIR="${OUTPUT_DIR:-../../models}"
 TYPE="${OUTTYPE:-f16}"
 METADATA_OVERRIDE="${METADATA_OVERRIDE:-}"
-CONVERTED_MODEL="${OUTPUT_DIR}/${MODEL_NAME}.gguf"
+if [[ -n "$MMPROJ" ]]; then
+    CONVERTED_MODEL="${OUTPUT_DIR}/mmproj-${MODEL_NAME}.gguf"
+else
+    CONVERTED_MODEL="${OUTPUT_DIR}/${MODEL_NAME}.gguf"
+fi
 
 echo "Model path: ${MODEL_PATH}"
 echo "Model name: ${MODEL_NAME}"
@@ -38,6 +42,7 @@ if [[ -n "$DEBUG" ]]; then
 else
     CMD_ARGS=("python")
 fi
+
 CMD_ARGS+=("../../convert_hf_to_gguf.py" "--verbose")
 CMD_ARGS+=("${MODEL_PATH}")
 CMD_ARGS+=("--outfile" "${CONVERTED_MODEL}")
@@ -50,7 +55,3 @@ CMD_ARGS+=("--outtype" "${TYPE}")
 echo ""
 echo "The environment variable CONVERTED_MODEL can be set to this path using:"
 echo "export CONVERTED_MODEL=$(realpath ${CONVERTED_MODEL})"
-if [[ -n "$MMPROJ" ]]; then
-    mmproj_file="${OUTPUT_DIR}/mmproj-$(basename "${CONVERTED_MODEL}")"
-    echo "The mmproj model was created in $(realpath "$mmproj_file")"
-fi

examples/parallel/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-parallel)
 add_executable(${TARGET} parallel.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/passkey/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-passkey)
 add_executable(${TARGET} passkey.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/retrieval/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-retrieval)
 add_executable(${TARGET} retrieval.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/save-load-state/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-save-load-state)
 add_executable(${TARGET} save-load-state.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/speculative-simple/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-speculative-simple)
 add_executable(${TARGET} speculative-simple.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/speculative-simple/speculative-simple.cpp

@@ -8,8 +8,24 @@
 #include <clocale>
 #include <cstdio>
 #include <cstring>
+#include <cinttypes>
 #include <string>
 #include <vector>
+#include <utility>
+
+struct spec_checkpoint {
+    int64_t n_tokens = 0;
+
+    std::vector<uint8_t> data;
+
+    size_t size() const {
+        return data.size();
+    }
+
+    bool empty() const {
+        return data.empty();
+    }
+};
 
 int main(int argc, char ** argv) {
     std::setlocale(LC_NUMERIC, "C");
@@ -27,7 +43,7 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
-    if (params.speculative.mparams_dft.path.empty()) {
+    if (params.speculative.draft.mparams.path.empty()) {
         LOG_ERR("%s: --model-draft is required\n", __func__);
         return 1;
     }
@@ -46,6 +62,14 @@ int main(int argc, char ** argv) {
     model_tgt = llama_init_tgt->model();
     ctx_tgt   = llama_init_tgt->context();
 
+    // check if the context supports partial sequence removal
+    const auto ctx_seq_rm = common_context_can_seq_rm(ctx_tgt);
+    const bool use_ckpt = (ctx_seq_rm == COMMON_CONTEXT_SEQ_RM_TYPE_FULL);
+
+    if (use_ckpt) {
+        LOG_INF("speculative decoding will use checkpoints (context does not support partial sequence removal)\n");
+    }
+
     const llama_vocab * vocab = llama_model_get_vocab(model_tgt);
 
     // load the draft model
@@ -53,7 +77,7 @@ int main(int argc, char ** argv) {
 
     // TODO: simplify this logic
     {
-        const auto & params_spec = params.speculative;
+        const auto & params_spec = params.speculative.draft;
 
         auto params_dft = params;
 
@@ -61,15 +85,15 @@ int main(int argc, char ** argv) {
         params_dft.n_ctx        = params_spec.n_ctx;
         params_dft.n_batch      = llama_n_ctx_seq(ctx_tgt);
         params_dft.devices      = params_spec.devices;
-        params_dft.model        = params_spec.mparams_dft;
+        params_dft.model        = params_spec.mparams;
         params_dft.n_gpu_layers = params_spec.n_gpu_layers;
 
         if (params_spec.cpuparams.n_threads > 0) {
-            params_dft.cpuparams.n_threads       = params.speculative.cpuparams.n_threads;
-            params_dft.cpuparams_batch.n_threads = params.speculative.cpuparams_batch.n_threads;
+            params_dft.cpuparams.n_threads       = params.speculative.draft.cpuparams.n_threads;
+            params_dft.cpuparams_batch.n_threads = params.speculative.draft.cpuparams_batch.n_threads;
         }
 
-        params_dft.tensor_buft_overrides = params.speculative.tensor_buft_overrides;
+        params_dft.tensor_buft_overrides = params.speculative.draft.tensor_buft_overrides;
 
         auto mparams_dft = common_model_params_to_llama(params_dft);
 
@@ -79,8 +103,8 @@ int main(int argc, char ** argv) {
             return 1;
         }
 
-        params.speculative.model_dft = model_dft.get();
-        params.speculative.cparams_dft = common_context_params_to_llama(params_dft);
+        params.speculative.draft.model = model_dft.get();
+        params.speculative.draft.cparams = common_context_params_to_llama(params_dft);
     }
 
     // Tokenize the prompt
@@ -119,7 +143,7 @@ int main(int argc, char ** argv) {
     const auto t_enc_start = ggml_time_us();
 
     // target model sampling context
-    struct common_sampler * smpl = common_sampler_init(model_tgt, params.sampling);
+    common_sampler_ptr smpl(common_sampler_init(model_tgt, params.sampling));
 
     // eval the prompt
     llama_decode(ctx_tgt, llama_batch_get_one(inp.data(), inp.size() - 1));
@@ -142,21 +166,49 @@ int main(int argc, char ** argv) {
 
     llama_batch batch_tgt = llama_batch_init(llama_n_batch(ctx_tgt), 0, 1);
 
+    size_t n_draft = 0;
+
+    llama_tokens draft;
+    spec_checkpoint spec_ckpt;
+
     const auto t_enc_end = ggml_time_us();
 
     const auto t_dec_start = ggml_time_us();
 
     while (true) {
-        // optionally, generate draft tokens that can be appended to the target batch
+        // generate or reuse draft tokens
         //
         // this is the most important part of the speculation. the more probable tokens that are provided here
         // the better the performance will be. in theory, this computation can be performed asynchronously and even
         // offloaded to a remote device. it doesn't even have to be based on an LLM. instead, it can provide tokens
         // from a cache or lookup tables.
         //
-        llama_tokens draft = common_speculative_draft(spec, params_spec, prompt_tgt, id_last);
+        if (draft.empty()) {
+            // generate a new draft
+            draft = common_speculative_draft(spec, params_spec, prompt_tgt, id_last);
 
-        //LOG_DBG("draft: %s\n", string_from(ctx_dft, draft).c_str());
+            // save the original draft size
+            n_draft = draft.size();
+
+            // save a checkpoint of the target context before evaluating the draft
+            // this allows us to restore the state if partial draft acceptance occurs
+            if (!draft.empty() && use_ckpt) {
+                const size_t ckpt_size = llama_state_seq_get_size_ext(ctx_tgt, 0, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);
+                spec_ckpt.data.resize(ckpt_size);
+
+                const size_t n = llama_state_seq_get_data_ext(ctx_tgt, spec_ckpt.data.data(), ckpt_size, 0, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);
+                GGML_ASSERT(n == ckpt_size);
+
+                spec_ckpt.n_tokens = (int64_t) prompt_tgt.size();
+                LOG_DBG("created speculative checkpoint (n_tokens = %" PRId64 ", size = %.3f MiB)\n",
+                        spec_ckpt.n_tokens, (float) spec_ckpt.data.size() / 1024 / 1024);
+            }
+        } else {
+            // we have a previous (partial) draft to reuse from checkpoint restoration
+            if (use_ckpt) {
+                GGML_ASSERT(!spec_ckpt.empty());
+            }
+        }
 
         // always have a token to evaluate from before - id_last
         common_batch_clear(batch_tgt);
@@ -164,11 +216,6 @@ int main(int argc, char ** argv) {
 
         // evaluate the target model on [id_last, draft0, draft1, ..., draftN-1]
         {
-            // do not waste time on small drafts
-            if (draft.size() < (size_t) params_spec.n_min) {
-                draft.clear();
-            }
-
             for (size_t i = 0; i < draft.size(); ++i) {
                 common_batch_add(batch_tgt, draft[i], n_past + i, { 0 }, true);
             }
@@ -178,6 +225,12 @@ int main(int argc, char ** argv) {
             llama_decode(ctx_tgt, batch_tgt);
         }
 
+        // only save the sampler sampler state if we use checkpoints
+        common_sampler_ptr smpl_save;
+        if (use_ckpt) {
+            smpl_save.reset(common_sampler_clone(smpl.get()));
+        }
+
         // sample from the full target batch and return the accepted tokens based on the target sampler
         //
         // for each token to be accepted, the sampler would have to sample that same token
@@ -185,14 +238,38 @@ int main(int argc, char ** argv) {
         // available logits from the batch and sample the next token until we run out of logits or the sampler
         // disagrees with the draft
         //
-        const auto ids = common_sampler_sample_and_accept_n(smpl, ctx_tgt, draft);
+        auto ids = common_sampler_sample_and_accept_n(smpl.get(), ctx_tgt, draft);
 
         //LOG_DBG("ids: %s\n", string_from(ctx_tgt, ids).c_str());
 
         GGML_ASSERT(ids.size() > 0); // there will always be at least one accepted token
 
+        // check for partial draft acceptance:
+        // if the context doesn't support partial sequence removal, restore the checkpoint
+        // and make the accepted tokens the new partial draft for the next iteration
+        if (use_ckpt && ids.size() - 1 < draft.size()) {
+            LOG_DBG("partial acceptance: %zu < %zu, restoring checkpoint\n", ids.size() - 1, draft.size());
+
+            draft = std::move(ids);
+
+            const size_t n = llama_state_seq_set_data_ext(ctx_tgt, spec_ckpt.data.data(), spec_ckpt.size(), 0, LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY);
+            GGML_ASSERT(n == spec_ckpt.size());
+
+            llama_memory_seq_rm(llama_get_memory(ctx_tgt), 0, spec_ckpt.n_tokens, -1);
+
+            prompt_tgt.resize(spec_ckpt.n_tokens);
+            smpl = std::move(smpl_save);
+
+            n_past = (int) prompt_tgt.size();
+
+            continue;
+        }
+
+        common_speculative_accept(spec, ids.size() - 1);
+
+        // full acceptance: consume the draft and commit accepted tokens
         n_past    += ids.size() - 1;
-        n_drafted += draft.size(); // note: we ignore the discarded small drafts
+        n_drafted += n_draft; // note: we ignore the discarded small drafts
         n_accept  += ids.size() - 1;
         n_predict += ids.size();
 
@@ -222,6 +299,9 @@ int main(int argc, char ** argv) {
 
         LOG_DBG("accepted %d/%d draft tokens, the last target token is: (%d)\n", (int) ids.size() - 1, (int) draft.size(), id_last);
 
+        // clear the draft since it has been consumed
+        draft.clear();
+
         {
             LOG_DBG("clear kv cache from any extra tokens, n_past = %d\n", n_past);
 
@@ -243,7 +323,7 @@ int main(int argc, char ** argv) {
     LOG_INF("decoded %4d tokens in %8.3f seconds, speed: %8.3f t/s\n", n_predict, (t_dec_end - t_dec_start) / 1e6f, n_predict  / ((t_dec_end - t_dec_start) / 1e6f));
 
     LOG_INF("\n");
-    LOG_INF("n_draft   = %d\n", params_spec.n_max);
+    LOG_INF("n_draft   = %d\n", params_spec.draft.n_max);
     LOG_INF("n_predict = %d\n", n_predict);
     LOG_INF("n_drafted = %d\n", n_drafted);
     LOG_INF("n_accept  = %d\n", n_accept);
@@ -254,11 +334,10 @@ int main(int argc, char ** argv) {
 
     LOG_INF("\n");
     LOG_INF("target:\n\n");
-    common_perf_print(ctx_tgt, smpl);
+    common_perf_print(ctx_tgt, smpl.get());
 
     llama_batch_free(batch_tgt);
 
-    common_sampler_free(smpl);
     common_speculative_free(spec);
 
     llama_backend_free();

examples/speculative/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-speculative)
 add_executable(${TARGET} speculative.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/speculative/speculative.cpp

@@ -49,7 +49,7 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
-    if (params.speculative.mparams_dft.path.empty()) {
+    if (params.speculative.draft.mparams.path.empty()) {
         LOG_ERR("%s: --model-draft is required\n", __func__);
         return 1;
     }
@@ -58,7 +58,7 @@ int main(int argc, char ** argv) {
     const int n_seq_dft = params.n_parallel;
 
     // probability threshold for splitting a draft branch (only for n_seq_dft > 1)
-    const float p_draft_split = params.speculative.p_split;
+    const float p_draft_split = params.speculative.draft.p_split;
 
     std::default_random_engine rng(params.sampling.seed == LLAMA_DEFAULT_SEED ? std::random_device()() : params.sampling.seed);
     std::uniform_real_distribution<> u_dist;
@@ -80,15 +80,15 @@ int main(int argc, char ** argv) {
     ctx_tgt   = llama_init_tgt->context();
 
     // load the draft model
-    params.devices = params.speculative.devices;
-    params.model = params.speculative.mparams_dft;
-    params.n_gpu_layers = params.speculative.n_gpu_layers;
-    if (params.speculative.cpuparams.n_threads > 0) {
-        params.cpuparams.n_threads = params.speculative.cpuparams.n_threads;
+    params.devices = params.speculative.draft.devices;
+    params.model = params.speculative.draft.mparams;
+    params.n_gpu_layers = params.speculative.draft.n_gpu_layers;
+    if (params.speculative.draft.cpuparams.n_threads > 0) {
+        params.cpuparams.n_threads = params.speculative.draft.cpuparams.n_threads;
     }
 
-    params.cpuparams_batch.n_threads = params.speculative.cpuparams_batch.n_threads;
-    params.tensor_buft_overrides     = params.speculative.tensor_buft_overrides;
+    params.cpuparams_batch.n_threads = params.speculative.draft.cpuparams_batch.n_threads;
+    params.tensor_buft_overrides     = params.speculative.draft.tensor_buft_overrides;
 
     auto llama_init_dft = common_init_from_params(params);
 
@@ -110,13 +110,21 @@ int main(int argc, char ** argv) {
         return 1;
     }
 
-    if (
-        llama_vocab_get_add_bos(vocab_tgt) != llama_vocab_get_add_bos(vocab_dft) ||
-        llama_vocab_get_add_eos(vocab_tgt) != llama_vocab_get_add_eos(vocab_dft) ||
-        llama_vocab_bos(vocab_tgt) != llama_vocab_bos(vocab_dft) ||
-        llama_vocab_eos(vocab_tgt) != llama_vocab_eos(vocab_dft)
-    ) {
-        LOG_ERR("%s: draft model special tokens must match target model to use speculation\n", __func__);
+    if (llama_vocab_get_add_bos(vocab_tgt) != llama_vocab_get_add_bos(vocab_dft) ||
+        (llama_vocab_get_add_bos(vocab_tgt) && llama_vocab_bos(vocab_tgt) != llama_vocab_bos(vocab_dft))) {
+        LOG_ERR("%s: draft model bos tokens must match target model to use speculation. add: %d - %d, id: %d - %d)\n",
+                __func__,
+                llama_vocab_get_add_bos(vocab_tgt), llama_vocab_get_add_bos(vocab_dft),
+                llama_vocab_bos(vocab_tgt), llama_vocab_bos(vocab_dft));
+        return 1;
+    }
+
+    if (llama_vocab_get_add_eos(vocab_tgt) != llama_vocab_get_add_eos(vocab_dft) ||
+        (llama_vocab_get_add_eos(vocab_tgt) && llama_vocab_eos(vocab_tgt) != llama_vocab_eos(vocab_dft))) {
+        LOG_ERR("%s: draft model eos tokens must match target model to use speculation. add: %d - %d, id: %d - %d)\n",
+                __func__,
+                llama_vocab_get_add_eos(vocab_tgt), llama_vocab_get_add_eos(vocab_dft),
+                llama_vocab_eos(vocab_tgt), llama_vocab_eos(vocab_dft));
         return 1;
     }
 
@@ -137,11 +145,12 @@ int main(int argc, char ** argv) {
         for (int i = SPEC_VOCAB_CHECK_START_TOKEN_ID; i < std::min(n_vocab_tgt, n_vocab_dft); ++i) {
             const char * token_text_tgt = llama_vocab_get_text(vocab_tgt, i);
             const char * token_text_dft = llama_vocab_get_text(vocab_dft, i);
+
             if (std::strcmp(token_text_tgt, token_text_dft) != 0) {
                 LOG_ERR("%s: draft model vocab must match target model to use speculation but ", __func__);
                 LOG_ERR("token %d content differs - target '%s', draft '%s'\n", i,
-                        common_token_to_piece(ctx_tgt, i).c_str(),
-                        common_token_to_piece(ctx_dft, i).c_str());
+                        common_token_to_piece(vocab_tgt, i).c_str(),
+                        common_token_to_piece(vocab_dft, i).c_str());
                 return 1;
             }
         }
@@ -183,7 +192,7 @@ int main(int argc, char ** argv) {
     //GGML_ASSERT(n_vocab == llama_vocab_n_tokens(model_dft));
 
     // how many tokens to draft each time
-    int n_draft = params.speculative.n_max;
+    int n_draft = params.speculative.draft.n_max;
 
     int n_predict = 0;
     int n_drafted = 0;

examples/sycl/CMakeLists.txt

@@ -5,5 +5,5 @@
 set(TARGET llama-ls-sycl-device)
 add_executable(${TARGET} ls-sycl-device.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/sycl/start-svr.sh

@@ -0,0 +1,124 @@
+#!/bin/bash
+
+#  MIT license
+#  Copyright (C) 2024 Intel Corporation
+#  SPDX-License-Identifier: MIT
+
+Help() {
+  cat << EOF
+Usage: $(basename "$0") [OPTIONS]
+
+This script processes files with specified options.
+
+Options:
+  -h, --help    Display this help message and exit.
+  -c, --context <value>    Set context length. Bigger need more memory.
+  -p, --promote <value>    Prompt to start generation with.
+  -m, --model   <value>    Full model file path.
+  -mg,--main-gpu <value>   Set main GPU ID (0 - n) for single GPU mode.
+  -sm,--split-mode <value> How to split the model across multiple GPUs, one of:
+                            - none: use one GPU only
+                            - layer (default): split layers and KV across GPUs
+                            - row: split rows across GPUs
+  -ngl,--n-gpu-layers <value>  Max. number of layers to store in VRAM (default: -1)
+  -lv,--log-verbosity <value>  Set the verbosity threshold. Messages with a higher verbosity will be
+                               ignored. Values:
+                                - 0: generic output
+                                - 1: error
+                                - 2: warning
+                                - 3: info
+                                - 4: debug
+
+
+EOF
+}
+
+BIN_FILE=./build/bin/llama-server
+SEED=0
+GPUS_SETTING=""
+
+MODEL_FILE=../models/Qwen3.5-4B-Q4_0.gguf
+NGL=99
+CONTEXT=4096
+GGML_SYCL_DEVICE=-1
+SPLIT_MODE=layer
+LOG_VERBOSE=3
+while [[ $# -gt 0 ]]; do
+    case "$1" in
+        -c|--context)
+            CONTEXT=$2
+            # Shift twice to consume both the option flag and its value
+            shift
+            shift
+            ;;
+        -m|--model)
+            MODEL_FILE="$2"
+            # Shift twice to consume both the option flag and its value
+            shift
+            shift
+            ;;
+        -mg|--main-gpu)
+            GGML_SYCL_DEVICE=$2
+            SPLIT_MODE=none
+            # Shift twice to consume both the option flag and its value
+            shift
+            shift
+            ;;
+        -sm|--split-mode)
+            SPLIT_MODE=$2
+            # Shift twice to consume both the option flag and its value
+            shift
+            shift
+            ;;
+        -ngl|--n-gpu-layers)
+            NGL=$2
+            # Shift twice to consume both the option flag and its value
+            shift
+            shift
+            ;;
+        -lv|--log-verbosity)
+            LOG_VERBOSE=$2
+            # Shift twice to consume both the option flag and its value
+            shift
+            shift
+            ;;
+        -h|--help)
+            Help
+            exit 0
+            ;;
+        *)
+            # Handle unknown options or stop processing options
+            echo "Invalid option: $1"
+            # Optional: exit script or shift to treat remaining as positional args
+            exit 1
+            ;;
+    esac
+done
+
+
+
+source /opt/intel/oneapi/setvars.sh
+
+#export GGML_SYCL_DEBUG=1
+
+#ZES_ENABLE_SYSMAN=1, Support to get free memory of GPU by sycl::aspect::ext_intel_free_memory. Recommended to use when --split-mode = layer.
+
+#support malloc device memory more than 4GB.
+export UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1
+echo "UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=${UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS}"
+
+if [ $GGML_SYCL_DEVICE -ne -1 ]; then
+    echo "Use $GGML_SYCL_DEVICE as main GPU"
+    #use signle GPU only
+    GPUS_SETTING="-mg $GGML_SYCL_DEVICE -sm ${SPLIT_MODE}"
+    export ONEAPI_DEVICE_SELECTOR="level_zero:${$GGML_SYCL_DEVICE}"
+    echo "ONEAPI_DEVICE_SELECTOR=${ONEAPI_DEVICE_SELECTOR}"
+else
+    echo "Use all Intel GPUs, including iGPU & dGPU"
+    GPUS_SETTING="-sm ${SPLIT_MODE}"
+ fi
+
+echo "run cmd: ZES_ENABLE_SYSMAN=1 ${BIN_FILE} -m ${MODEL_FILE} -no-cnv -p "${INPUT_PROMPT}" -n 200 -e -ngl ${NGL} -s ${SEED} -c ${CONTEXT} ${GPUS_SETTING} -lv ${LOG_VERBOSE}  --mmap "
+ZES_ENABLE_SYSMAN=1 ${BIN_FILE} -m ${MODEL_FILE} -ngl ${NGL} -s ${SEED} -c ${CONTEXT} ${GPUS_SETTING} -lv ${LOG_VERBOSE} --mmap --host 0.0.0.0 --port 8000
+
+

examples/sycl/test.sh

@@ -38,7 +38,7 @@ SEED=0
 GPUS_SETTING=""
 
 INPUT_PROMPT="Building a website can be done in 10 simple steps:\nStep 1:"
-MODEL_FILE=models/llama-2-7b.Q4_0.gguf
+MODEL_FILE=../models/llama-2-7b.Q4_0.gguf
 NGL=99
 CONTEXT=4096
 GGML_SYCL_DEVICE=-1
@@ -122,9 +122,10 @@ if [ $GGML_SYCL_DEVICE -ne -1 ]; then
     export ONEAPI_DEVICE_SELECTOR="level_zero:${$GGML_SYCL_DEVICE}"
     echo "ONEAPI_DEVICE_SELECTOR=${ONEAPI_DEVICE_SELECTOR}"
 else
-   echo "Use all Intel GPUs, including iGPU & dGPU"
+    echo "Use all Intel GPUs, including iGPU & dGPU"
+    GPUS_SETTING="-sm ${SPLIT_MODE}"
  fi
 
-echo "run cmd: ZES_ENABLE_SYSMAN=1 ${BIN_FILE} -m ${MODEL_FILE} -no-cnv -p "${INPUT_PROMPT}" -n 400 -e -ngl ${NGL} -s ${SEED} -c ${CONTEXT} ${GPUS_SETTING} -lv ${LOG_VERBOSE}  --mmap "
-ZES_ENABLE_SYSMAN=1 ${BIN_FILE} -m ${MODEL_FILE} -no-cnv -p "${INPUT_PROMPT}" -n 400 -e -ngl ${NGL} -s ${SEED} -c ${CONTEXT} ${GPUS_SETTING} -lv ${LOG_VERBOSE} --mmap
+echo "run cmd: ZES_ENABLE_SYSMAN=1 ${BIN_FILE} -m ${MODEL_FILE} -no-cnv -p "${INPUT_PROMPT}" -n 200 -e -ngl ${NGL} -s ${SEED} -c ${CONTEXT} ${GPUS_SETTING} -lv ${LOG_VERBOSE}  --mmap "
+ZES_ENABLE_SYSMAN=1 ${BIN_FILE} -m ${MODEL_FILE} -no-cnv -p "${INPUT_PROMPT}" -n 200 -e -ngl ${NGL} -s ${SEED} -c ${CONTEXT} ${GPUS_SETTING} -lv ${LOG_VERBOSE} --mmap
 

examples/sycl/win-start-svr.bat

@@ -0,0 +1,179 @@
+::  MIT license
+::  Copyright (C) 2024 Intel Corporation
+::  SPDX-License-Identifier: MIT
+
+@echo off
+setlocal EnableExtensions EnableDelayedExpansion
+
+set "BIN_FILE=.\build\bin\llama-server.exe"
+set "SEED=0"
+set "GPUS_SETTING="
+
+set "MODEL_FILE=..\models\Qwen3.5-4B-Q4_0.gguf"
+set "NGL=99"
+set "CONTEXT=4096"
+set "GGML_SYCL_DEVICE=-1"
+set "SPLIT_MODE=layer"
+set "LOG_VERBOSE=3"
+
+if "%~1"=="" goto after_args
+
+:parse_args
+if "%~1"=="" goto after_args
+
+if /I "%~1"=="-c" (
+  if "%~2"=="" goto missing_value
+  set "CONTEXT=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--context" (
+  if "%~2"=="" goto missing_value
+  set "CONTEXT=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-m" (
+  if "%~2"=="" goto missing_value
+  set "MODEL_FILE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--model" (
+  if "%~2"=="" goto missing_value
+  set "MODEL_FILE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-mg" (
+  if "%~2"=="" goto missing_value
+  set "GGML_SYCL_DEVICE=%~2"
+  set "SPLIT_MODE=none"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--main-gpu" (
+  if "%~2"=="" goto missing_value
+  set "GGML_SYCL_DEVICE=%~2"
+  set "SPLIT_MODE=none"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-sm" (
+  if "%~2"=="" goto missing_value
+  set "SPLIT_MODE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--split-mode" (
+  if "%~2"=="" goto missing_value
+  set "SPLIT_MODE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-ngl" (
+  if "%~2"=="" goto missing_value
+  set "NGL=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--n-gpu-layers" (
+  if "%~2"=="" goto missing_value
+  set "NGL=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-lv" (
+  if "%~2"=="" goto missing_value
+  set "LOG_VERBOSE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--log-verbosity" (
+  if "%~2"=="" goto missing_value
+  set "LOG_VERBOSE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-h" goto help
+if /I "%~1"=="--help" goto help
+
+echo Invalid option: %~1
+exit /b 1
+
+:missing_value
+echo Missing value for option: %~1
+exit /b 1
+
+:help
+echo Usage: %~n0 [OPTIONS]
+echo.
+echo This script processes files with specified options.
+echo.
+echo Options:
+echo   -h, --help    Display this help message and exit.
+echo   -c, --context ^<value^>    Set context length. Bigger need more memory.
+echo   -m, --model   ^<value^>    Full model file path.
+echo   -mg,--main-gpu ^<value^>   Set main GPU ID (0 - n) for single GPU mode.
+echo   -sm,--split-mode ^<value^> How to split the model across multiple GPUs, one of:
+echo                             - none: use one GPU only
+echo                             - layer (default): split layers and KV across GPUs
+echo                             - row: split rows across GPUs
+echo   -ngl,--n-gpu-layers ^<value^>  Max. number of layers to store in VRAM (default: -1)
+echo   -lv,--log-verbosity ^<value^>  Set the verbosity threshold. Messages with a higher verbosity will be
+echo                                ignored. Values:
+echo                                 - 0: generic output
+echo                                 - 1: error
+echo                                 - 2: warning
+echo                                 - 3: info
+echo                                 - 4: debug
+exit /b 0
+
+:after_args
+
+REM In Windows CMD, source is not available; call oneAPI setvars if present.
+if exist "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" (
+  call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" >nul
+) else (
+  echo Warning: oneAPI setvars.bat not found. Continuing without environment setup.
+)
+
+REM Support malloc device memory more than 4GB.
+set "UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1"
+echo UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=%UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS%
+
+if not "%GGML_SYCL_DEVICE%"=="-1" (
+  echo Use %GGML_SYCL_DEVICE% as main GPU
+  REM Use single GPU only.
+  set "GPUS_SETTING=-mg %GGML_SYCL_DEVICE% -sm %SPLIT_MODE%"
+  set "ONEAPI_DEVICE_SELECTOR=level_zero:%GGML_SYCL_DEVICE%"
+  echo ONEAPI_DEVICE_SELECTOR=%ONEAPI_DEVICE_SELECTOR%
+) else (
+  echo Use all Intel GPUs, including iGPU ^& dGPU
+  set "GPUS_SETTING=-sm %SPLIT_MODE%"
+)
+
+echo run cmd: ZES_ENABLE_SYSMAN=1 %BIN_FILE% -m "%MODEL_FILE%" -ngl %NGL% -s %SEED% -c %CONTEXT% %GPUS_SETTING% -lv %LOG_VERBOSE% --mmap --host 0.0.0.0 --port 8000
+set "ZES_ENABLE_SYSMAN=1"
+%BIN_FILE% -m "%MODEL_FILE%" -ngl %NGL% -s %SEED% -c %CONTEXT% %GPUS_SETTING% -lv %LOG_VERBOSE% --mmap --host 0.0.0.0 --port 8000
+
+endlocal
+

examples/sycl/win-test.bat

@@ -2,10 +2,200 @@
 ::  Copyright (C) 2024 Intel Corporation
 ::  SPDX-License-Identifier: MIT
 
-set INPUT2="Building a website can be done in 10 simple steps:\nStep 1:"
-@call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
 
-:: support malloc device memory more than 4GB.
-set UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1
-set LOAD_MODE="--mmap"
-.\build\bin\llama-completion.exe -m models\llama-2-7b.Q4_0.gguf -no-cnv -p %INPUT2% -n 400 -e -ngl 99 -s 0 %LOAD_MODE%
+@echo off
+setlocal EnableExtensions EnableDelayedExpansion
+
+REM MIT license
+REM Copyright (C) 2024 Intel Corporation
+REM SPDX-License-Identifier: MIT
+
+set "BIN_FILE=.\build\bin\llama-completion.exe"
+set "SEED=0"
+set "GPUS_SETTING="
+
+set "INPUT_PROMPT=Building a website can be done in 10 simple steps:^nStep 1:"
+set "MODEL_FILE=..\models\llama-2-7b.Q4_0.gguf"
+set "NGL=99"
+set "CONTEXT=4096"
+set "GGML_SYCL_DEVICE=-1"
+set "SPLIT_MODE=layer"
+set "LOG_VERBOSE=3"
+
+if "%~1"=="" goto after_args
+
+:parse_args
+if "%~1"=="" goto after_args
+
+if /I "%~1"=="-c" (
+  if "%~2"=="" goto missing_value
+  set "CONTEXT=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--context" (
+  if "%~2"=="" goto missing_value
+  set "CONTEXT=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-p" (
+  if "%~2"=="" goto missing_value
+  set "INPUT_PROMPT=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--promote" (
+  if "%~2"=="" goto missing_value
+  set "INPUT_PROMPT=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-m" (
+  if "%~2"=="" goto missing_value
+  set "MODEL_FILE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--model" (
+  if "%~2"=="" goto missing_value
+  set "MODEL_FILE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-mg" (
+  if "%~2"=="" goto missing_value
+  set "GGML_SYCL_DEVICE=%~2"
+  set "SPLIT_MODE=none"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--main-gpu" (
+  if "%~2"=="" goto missing_value
+  set "GGML_SYCL_DEVICE=%~2"
+  set "SPLIT_MODE=none"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-sm" (
+  if "%~2"=="" goto missing_value
+  set "SPLIT_MODE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--split-mode" (
+  if "%~2"=="" goto missing_value
+  set "SPLIT_MODE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-ngl" (
+  if "%~2"=="" goto missing_value
+  set "NGL=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--n-gpu-layers" (
+  if "%~2"=="" goto missing_value
+  set "NGL=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-lv" (
+  if "%~2"=="" goto missing_value
+  set "LOG_VERBOSE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+if /I "%~1"=="--log-verbosity" (
+  if "%~2"=="" goto missing_value
+  set "LOG_VERBOSE=%~2"
+  shift
+  shift
+  goto parse_args
+)
+
+if /I "%~1"=="-h" goto help
+if /I "%~1"=="--help" goto help
+
+echo Invalid option: %~1
+exit /b 1
+
+:missing_value
+echo Missing value for option: %~1
+exit /b 1
+
+:help
+echo Usage: %~n0 [OPTIONS]
+echo.
+echo This script processes files with specified options.
+echo.
+echo Options:
+echo   -h, --help    Display this help message and exit.
+echo   -c, --context ^<value^>    Set context length. Bigger need more memory.
+echo   -p, --promote ^<value^>    Prompt to start generation with.
+echo   -m, --model   ^<value^>    Full model file path.
+echo   -mg,--main-gpu ^<value^>   Set main GPU ID (0 - n) for single GPU mode.
+echo   -sm,--split-mode ^<value^> How to split the model across multiple GPUs, one of:
+echo                             - none: use one GPU only
+echo                             - layer (default): split layers and KV across GPUs
+echo                             - row: split rows across GPUs
+echo   -ngl,--n-gpu-layers ^<value^>  Max. number of layers to store in VRAM (default: -1)
+echo   -lv,--log-verbosity ^<value^>  Set the verbosity threshold. Messages with a higher verbosity will be
+echo                                ignored. Values:
+echo                                 - 0: generic output
+echo                                 - 1: error
+echo                                 - 2: warning
+echo                                 - 3: info
+echo                                 - 4: debug
+exit /b 0
+
+:after_args
+
+REM In Windows CMD, source is not available; call oneAPI setvars if present.
+if exist "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" (
+  call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" >nul
+) else (
+  echo Warning: oneAPI setvars.bat not found. Continuing without environment setup.
+)
+
+REM Support malloc device memory more than 4GB.
+set "UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=1"
+echo UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS=%UR_L0_ENABLE_RELAXED_ALLOCATION_LIMITS%
+
+if not "%GGML_SYCL_DEVICE%"=="-1" (
+  echo Use %GGML_SYCL_DEVICE% as main GPU
+  REM Use single GPU only.
+  set "GPUS_SETTING=-mg %GGML_SYCL_DEVICE% -sm %SPLIT_MODE%"
+  set "ONEAPI_DEVICE_SELECTOR=level_zero:%GGML_SYCL_DEVICE%"
+  echo ONEAPI_DEVICE_SELECTOR=%ONEAPI_DEVICE_SELECTOR%
+) else (
+  echo Use all Intel GPUs, including iGPU ^& dGPU
+  set "GPUS_SETTING=-sm %SPLIT_MODE%"
+)
+
+echo run cmd: ZES_ENABLE_SYSMAN=1 %BIN_FILE% -m %MODEL_FILE% -no-cnv -p "%INPUT_PROMPT%" -n 200 -e -ngl %NGL% -s %SEED% -c %CONTEXT% %GPUS_SETTING% -lv %LOG_VERBOSE% --mmap
+set "ZES_ENABLE_SYSMAN=1"
+%BIN_FILE% -m "%MODEL_FILE%" -no-cnv -p "%INPUT_PROMPT%" -n 200 -e -ngl %NGL% -s %SEED% -c %CONTEXT% %GPUS_SETTING% -lv %LOG_VERBOSE% --mmap
+
+endlocal
+

examples/training/CMakeLists.txt

@@ -1,5 +1,5 @@
 set(TARGET llama-finetune)
 add_executable(${TARGET} finetune.cpp)
 install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_11)

ggml/CMakeLists.txt

@@ -1,17 +1,11 @@
 cmake_minimum_required(VERSION 3.14...3.28) # for add_link_options and implicit target directories.
 
-# ref: https://cmake.org/cmake/help/latest/policy/CMP0194.html
-# MSVC is not a valid assembler for the ASM language.
-# Set to NEW to avoid a warning on CMake 4.1+ with MSVC.
-if (POLICY CMP0194)
-    cmake_policy(SET CMP0194 NEW)
-endif()
 project("ggml" C CXX ASM)
 
 ### GGML Version
 set(GGML_VERSION_MAJOR 0)
-set(GGML_VERSION_MINOR 9)
-set(GGML_VERSION_PATCH 11)
+set(GGML_VERSION_MINOR 10)
+set(GGML_VERSION_PATCH 2)
 set(GGML_VERSION_BASE "${GGML_VERSION_MAJOR}.${GGML_VERSION_MINOR}.${GGML_VERSION_PATCH}")
 
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake/")
@@ -219,7 +213,7 @@ set   (GGML_CUDA_COMPRESSION_MODE "size" CACHE STRING
 set_property(CACHE GGML_CUDA_COMPRESSION_MODE PROPERTY STRINGS "none;speed;balance;size")
 
 option(GGML_HIP                             "ggml: use HIP"                                   OFF)
-option(GGML_HIP_GRAPHS                      "ggml: use HIP graph, experimental, slow"         OFF)
+option(GGML_HIP_GRAPHS                      "ggml: use HIP graph"                              ON)
 option(GGML_HIP_RCCL                        "ggml: use ROCm Collective Comm. Library"         OFF)
 option(GGML_HIP_NO_VMM                      "ggml: do not try to use HIP VMM"                 ON)
 option(GGML_HIP_ROCWMMA_FATTN               "ggml: enable rocWMMA for FlashAttention"         OFF)
@@ -254,6 +248,7 @@ option(GGML_RPC                             "ggml: use RPC"
 option(GGML_SYCL                            "ggml: use SYCL"                                  OFF)
 option(GGML_SYCL_F16                        "ggml: use 16 bit floats for sycl calculations"   OFF)
 option(GGML_SYCL_GRAPH                      "ggml: enable graphs in the SYCL backend"         ON)
+option(GGML_SYCL_HOST_MEM_FALLBACK          "ggml: allow host memory fallback in SYCL reorder (requires kernel 6.8+)" ON)
 option(GGML_SYCL_DNN                        "ggml: enable oneDNN in the SYCL backend"         ON)
 set   (GGML_SYCL_TARGET "INTEL" CACHE STRING
                                             "ggml: sycl target device")

ggml/src/CMakeLists.txt

@@ -470,11 +470,10 @@ endforeach()
 
 target_link_libraries(ggml-base PRIVATE Threads::Threads)
 
-find_library(MATH_LIBRARY m)
-if (MATH_LIBRARY)
-    if (NOT WIN32 OR NOT DEFINED ENV{ONEAPI_ROOT})
-        target_link_libraries(ggml-base PRIVATE m)
-    endif()
+if (DEFINED MATH_LIBRARY)
+    target_link_libraries(ggml-base PRIVATE ${MATH_LIBRARY})
+elseif (NOT WIN32 AND NOT DEFINED ENV{ONEAPI_ROOT})
+    target_link_libraries(ggml-base PRIVATE m)
 endif()
 
 if (CMAKE_SYSTEM_NAME MATCHES "Android")

ggml/src/ggml-backend-meta.cpp

@@ -1133,7 +1133,7 @@ static enum ggml_status ggml_backend_meta_buffer_init_tensor(ggml_backend_buffer
         if (t_ij->view_src != nullptr && ggml_backend_buffer_is_meta(t_ij->view_src->buffer)) {
             t_ij->view_src = ggml_backend_meta_buffer_simple_tensor(tensor->view_src, j);
             if (t_ij->view_offs > 0 && split_dim >= 0 && split_dim < GGML_MAX_DIMS) {
-                GGML_ASSERT(ne[split_dim] != 0 && tensor->ne[split_dim] != 0);
+                GGML_ASSERT(tensor->ne[split_dim] != 0);
                 const int split_dim_view_src = ggml_backend_meta_get_split_state(tensor->view_src, /*assume_sync =*/ true).axis;
                 GGML_ASSERT(split_dim_view_src >= 0 && split_dim_view_src < GGML_MAX_DIMS);
 
@@ -1170,6 +1170,28 @@ static enum ggml_status ggml_backend_meta_buffer_init_tensor(ggml_backend_buffer
 
         simple_tensors.push_back(t_ij);
     }
+
+    // If one of the sources has a zero-sized slice, disable the computation:
+    for (int i = 0; i < GGML_MAX_SRC; i++) {
+        if (tensor->src[i] == nullptr || !ggml_backend_buffer_is_meta(tensor->src[i]->buffer)) {
+            continue;
+        }
+
+        const ggml_backend_meta_split_state split_state_src = ggml_backend_meta_get_split_state(tensor->src[i], /*assume_sync =*/ true);
+        if (split_state_src.axis < 0 || split_state_src.axis >= GGML_MAX_DIMS) {
+            continue;
+        }
+        for (size_t j = 0; j < n_simple_bufs; j++) {
+            int64_t ne_sum = 0;
+            for (size_t s = 0; s < split_state_src.n_segments; s++) {
+                ne_sum += split_state_src.ne[s*n_simple_bufs + j];
+            }
+            if (ne_sum == 0) {
+                simple_tensors[j]->flags &= ~GGML_TENSOR_FLAG_COMPUTE;
+            }
+        }
+    }
+
     buf_ctx->simple_tensors[tensor] = simple_tensors;
 
     return GGML_STATUS_SUCCESS;
@@ -1183,40 +1205,57 @@ static void ggml_backend_meta_buffer_set_tensor(ggml_backend_buffer_t buffer, gg
 
     if (split_state.n_segments != 1) {
         GGML_ASSERT(split_state.axis >= 0 && split_state.axis < GGML_MAX_DIMS);
-        GGML_ASSERT(offset == 0);
-        GGML_ASSERT(size == ggml_nbytes(tensor));
         GGML_ASSERT(tensor->ne[3] == 1);
+
         size_t offset_data = 0;
         std::vector<size_t> simple_offsets(n_bufs, 0);
         if (split_state.axis == GGML_BACKEND_SPLIT_AXIS_0) {
             GGML_ASSERT(tensor->ne[2] == 1);
+
+            const size_t row_stride = tensor->nb[1];
+            GGML_ASSERT(offset % row_stride == 0);
+            GGML_ASSERT(size   % row_stride == 0);
+            const int64_t r_start = offset / row_stride;
+            const int64_t r_count = size   / row_stride;
+            GGML_ASSERT(r_start + r_count <= tensor->ne[1]);
+
             const int64_t blck_size = ggml_blck_size(tensor->type);
             for (size_t s = 0; s < split_state.n_segments; s++) {
                 for (size_t j = 0; j < n_bufs; j++) {
                     ggml_tensor * simple_tensor = ggml_backend_meta_buffer_simple_tensor(tensor, j);
                     GGML_ASSERT(split_state.ne[s*n_bufs + j] % blck_size == 0);
                     const size_t nbytes = split_state.ne[s*n_bufs + j]/blck_size * tensor->nb[0];
-                    ggml_backend_tensor_set_2d(simple_tensor, (const char *) data + offset_data, simple_offsets[j], nbytes,
-                        tensor->ne[1], simple_tensor->nb[1], tensor->nb[1]);
+                    ggml_backend_tensor_set_2d(simple_tensor, (const char *) data + offset_data,
+                        simple_offsets[j] + r_start * simple_tensor->nb[1], nbytes,
+                        r_count, simple_tensor->nb[1], tensor->nb[1]);
                     offset_data       += nbytes;
                     simple_offsets[j] += nbytes;
                 }
             }
-            GGML_ASSERT(offset_data*tensor->ne[1] == size);
+            GGML_ASSERT(offset_data*r_count == size);
             return;
         }
         GGML_ASSERT(split_state.axis == GGML_BACKEND_SPLIT_AXIS_1);
+
+        const size_t row_stride = tensor->nb[2];
+        GGML_ASSERT(offset % row_stride == 0);
+        GGML_ASSERT(size   % row_stride == 0);
+        const int64_t r_start = offset / row_stride;
+        const int64_t r_count = size   / row_stride;
+        GGML_ASSERT(r_start + r_count <= tensor->ne[2]);
+
         for (size_t s = 0; s < split_state.n_segments; s++) {
             for (size_t j = 0; j < n_bufs; j++) {
                 ggml_tensor * simple_tensor = ggml_backend_meta_buffer_simple_tensor(tensor, j);
                 const size_t nbytes = split_state.ne[s*n_bufs + j] * tensor->nb[1];
-                ggml_backend_tensor_set_2d(simple_tensor, (const char *) data + offset_data, simple_offsets[j], nbytes,
-                    tensor->ne[2], simple_tensor->nb[2], tensor->nb[2]);
+                ggml_backend_tensor_set_2d(simple_tensor, (const char *) data + offset_data,
+                    simple_offsets[j] + r_start * simple_tensor->nb[2], nbytes,
+                    r_count, simple_tensor->nb[2], tensor->nb[2]);
                 offset_data       += nbytes;
                 simple_offsets[j] += nbytes;
             }
         }
-        GGML_ASSERT(offset_data*tensor->ne[2] == size);
+        GGML_ASSERT(offset_data*r_count == size);
         return;
     }
 
@@ -1270,7 +1309,62 @@ static void ggml_backend_meta_buffer_get_tensor(ggml_backend_buffer_t buffer, co
     GGML_ASSERT(ggml_is_contiguous(tensor));
 
     const ggml_backend_meta_split_state split_state = ggml_backend_meta_get_split_state(tensor, /*assume_sync =*/ false);
-    GGML_ASSERT(split_state.n_segments == 1);
+
+    if (split_state.n_segments != 1) {
+        GGML_ASSERT(split_state.axis >= 0 && split_state.axis < GGML_MAX_DIMS);
+        GGML_ASSERT(tensor->ne[3] == 1);
+
+        size_t offset_data = 0;
+        std::vector<size_t> simple_offsets(n_bufs, 0);
+        if (split_state.axis == GGML_BACKEND_SPLIT_AXIS_0) {
+            GGML_ASSERT(tensor->ne[2] == 1);
+
+            const size_t row_stride = tensor->nb[1];
+            GGML_ASSERT(offset % row_stride == 0);
+            GGML_ASSERT(size   % row_stride == 0);
+            const int64_t r_start = offset / row_stride;
+            const int64_t r_count = size   / row_stride;
+            GGML_ASSERT(r_start + r_count <= tensor->ne[1]);
+
+            const int64_t blck_size = ggml_blck_size(tensor->type);
+            for (size_t s = 0; s < split_state.n_segments; s++) {
+                for (size_t j = 0; j < n_bufs; j++) {
+                    const ggml_tensor * simple_tensor = ggml_backend_meta_buffer_simple_tensor(tensor, j);
+                    GGML_ASSERT(split_state.ne[s*n_bufs + j] % blck_size == 0);
+                    const size_t nbytes = split_state.ne[s*n_bufs + j]/blck_size * tensor->nb[0];
+                    ggml_backend_tensor_get_2d(simple_tensor, (char *) data + offset_data,
+                        simple_offsets[j] + r_start * simple_tensor->nb[1], nbytes,
+                        r_count, simple_tensor->nb[1], tensor->nb[1]);
+                    offset_data       += nbytes;
+                    simple_offsets[j] += nbytes;
+                }
+            }
+            GGML_ASSERT(offset_data*r_count == size);
+            return;
+        }
+        GGML_ASSERT(split_state.axis == GGML_BACKEND_SPLIT_AXIS_1);
+
+        const size_t row_stride = tensor->nb[2];
+        GGML_ASSERT(offset % row_stride == 0);
+        GGML_ASSERT(size   % row_stride == 0);
+        const int64_t r_start = offset / row_stride;
+        const int64_t r_count = size   / row_stride;
+        GGML_ASSERT(r_start + r_count <= tensor->ne[2]);
+
+        for (size_t s = 0; s < split_state.n_segments; s++) {
+            for (size_t j = 0; j < n_bufs; j++) {
+                const ggml_tensor * simple_tensor = ggml_backend_meta_buffer_simple_tensor(tensor, j);
+                const size_t nbytes = split_state.ne[s*n_bufs + j] * tensor->nb[1];
+                ggml_backend_tensor_get_2d(simple_tensor, (char *) data + offset_data,
+                    simple_offsets[j] + r_start * simple_tensor->nb[2], nbytes,
+                    r_count, simple_tensor->nb[2], tensor->nb[2]);
+                offset_data       += nbytes;
+                simple_offsets[j] += nbytes;
+            }
+        }
+        GGML_ASSERT(offset_data*r_count == size);
+        return;
+    }
 
     switch (split_state.axis) {
         case GGML_BACKEND_SPLIT_AXIS_0:
@@ -1404,26 +1498,32 @@ struct ggml_backend_meta_context {
     struct backend_config {
         ggml_backend_t backend;
 
-        std::vector<cgraph_config> cgraphs;
-        std::vector<ggml_tensor *> nodes;
-        ggml_backend_buffer_ptr    buf;
+        std::vector<cgraph_config>           cgraphs;
+        std::vector<ggml_tensor *>           nodes;
+        std::vector<ggml_backend_buffer_ptr> bufs;
 
-        backend_config(ggml_backend_t backend) : backend(backend) {}
+        backend_config(ggml_backend_t backend, const size_t n_reduce_steps) : backend(backend) {
+            bufs.resize(n_reduce_steps);
+        }
     };
     std::string                 name;
     std::vector<backend_config> backend_configs;
     ggml_context_ptr            ctx;
     std::vector<ggml_cgraph *>  cgraphs_aux;
     std::vector<ggml_tensor *>  nodes_aux;
+    size_t                      n_reduce_steps;
     int                         max_nnodes    = 0;
     size_t                      max_tmp_size  = 0;
     size_t                      max_subgraphs = 0;
+    size_t                      n_subgraphs   = 0;
+    uint64_t                    uid           = 0;
 
     void *                               comm_ctx       = nullptr;
     ggml_backend_comm_allreduce_tensor_t comm_allreduce = nullptr;
 
     ggml_backend_meta_context(ggml_backend_dev_t meta_dev, const char * params) {
         const size_t n_devs = ggml_backend_meta_dev_n_devs(meta_dev);
+        n_reduce_steps = std::ceil(std::log2(n_devs));
         name = "Meta(";
         std::vector<ggml_backend_t> simple_backends;
         backend_configs.reserve(n_devs);
@@ -1435,7 +1535,7 @@ struct ggml_backend_meta_context {
             }
             name += ggml_backend_dev_name(simple_dev);
             simple_backends.push_back(ggml_backend_dev_init(simple_dev, params));
-            backend_configs.emplace_back(simple_backends.back());
+            backend_configs.emplace_back(simple_backends.back(), n_reduce_steps);
         }
         name += ")";
 
@@ -1465,10 +1565,6 @@ struct ggml_backend_meta_context {
             ggml_backend_free(bc.backend);
         }
     }
-
-    size_t n_reduce_steps() const {
-        return std::ceil(std::log2(backend_configs.size()));
-    }
 };
 
 static const char * ggml_backend_meta_get_name(ggml_backend_t backend) {
@@ -1578,6 +1674,9 @@ static enum ggml_status ggml_backend_meta_graph_compute(ggml_backend_t backend,
     const size_t n_backends = ggml_backend_meta_n_backends(backend);
     ggml_backend_meta_context * backend_ctx = (ggml_backend_meta_context *) backend->context;
 
+    // If the previous cgraph had a defined UID it can be used to skip rebuilding the subgraphs per simple backend.
+    const bool needs_rebuild = (cgraph->uid == 0) || (cgraph->uid != backend_ctx->uid);
+
     bool max_nnodes_raised = false;
     if (cgraph->n_nodes > backend_ctx->max_nnodes) {
         for (size_t j = 0; j < n_backends; j++) {
@@ -1587,173 +1686,233 @@ static enum ggml_status ggml_backend_meta_graph_compute(ggml_backend_t backend,
         }
         backend_ctx->max_nnodes = cgraph->n_nodes;
         max_nnodes_raised = true;
+        assert(needs_rebuild);
     }
-    for (size_t j = 0; j < n_backends; j++) {
-        auto & bcj = backend_ctx->backend_configs[j];
 
-        for (int i = 0; i < cgraph->n_nodes; i++) {
-            ggml_tensor * node = cgraph->nodes[i];
-            if (node->view_src != nullptr && node->view_src->op == GGML_OP_NONE && ggml_backend_buffer_is_host(node->view_src->buffer)) {
-                // FIXME s_copy_main is on the CPU and its view seems to be incorrectly added to the graph nodes.
-                // For regular usage this doesn't matter since it's a noop but trying to call ggml_backend_meta_buffer_simple_tensor results in a crash.
-                bcj.nodes[i] = node;
-                continue;
+    if (needs_rebuild) {
+        size_t n_subgraphs  = 0;
+        size_t max_tmp_size = 0;
+
+        for (size_t j = 0; j < n_backends; j++) {
+            auto & bcj = backend_ctx->backend_configs[j];
+
+            for (int i = 0; i < cgraph->n_nodes; i++) {
+                ggml_tensor * node = cgraph->nodes[i];
+                if (node->view_src != nullptr && node->view_src->op == GGML_OP_NONE && ggml_backend_buffer_is_host(node->view_src->buffer)) {
+                    // FIXME s_copy_main is on the CPU and its view seems to be incorrectly added to the graph nodes.
+                    // For regular usage this doesn't matter since it's a noop but trying to call ggml_backend_meta_buffer_simple_tensor results in a crash.
+                    bcj.nodes[i] = node;
+                    continue;
+                }
+                bcj.nodes[i] = ggml_backend_meta_buffer_simple_tensor(node, j);
+                GGML_ASSERT(bcj.nodes[i]);
             }
-            bcj.nodes[i] = ggml_backend_meta_buffer_simple_tensor(node, j);
-            GGML_ASSERT(bcj.nodes[i]);
         }
-    }
 
-    size_t n_subgraphs  = 0;
-    size_t max_tmp_size = 0;
-    {
-        // For MoE models it may make sense to delay the AllReduce in order to reduce I/O:
-        auto get_i_delayed = [&](const int i) -> int {
-            int id = i; // i_delayed
-            int idr = i; // i_delayed return, last safe return value
+        {
+            // For MoE models it may make sense to delay the AllReduce in order to reduce I/O:
+            auto get_i_delayed = [&](const int i) -> int {
+                int id = i; // i_delayed
+                int idr = i; // i_delayed return, last safe return value
 
-            ggml_tensor * node = cgraph->nodes[id];
-            int32_t n_used = ggml_node_get_use_count(cgraph, id);
-            if (id + 1 >= cgraph->n_nodes) {
-                return idr;
-            }
-            {
-                ggml_tensor * next = cgraph->nodes[id+1];
-                if (next->op == GGML_OP_ADD_ID && next->src[0] == node &&
-                        ggml_backend_meta_get_split_state(next->src[1], false).axis == GGML_BACKEND_SPLIT_AXIS_PARTIAL &&
-                        ggml_backend_meta_get_split_state(next->src[2], false).axis == GGML_BACKEND_SPLIT_AXIS_MIRRORED) {
-                    node = next;
+                ggml_tensor * node = cgraph->nodes[id];
+                int32_t n_used = ggml_node_get_use_count(cgraph, id);
+
+                // Skip MIRRORED nodes that don't consume node
+                auto skip_unrelated = [&]() {
+                    while (id + 1 < cgraph->n_nodes) {
+                        ggml_tensor * next = cgraph->nodes[id+1];
+                        if (ggml_backend_meta_get_split_state(next, false).axis != GGML_BACKEND_SPLIT_AXIS_MIRRORED) {
+                            break;
+                        }
+                        bool safe = true;
+                        for (int s = 0; s < GGML_MAX_SRC; s++) {
+                            if (next->src[s] == nullptr) {
+                                continue;
+                            }
+                            if (next->src[s] == node) {
+                                safe = false;
+                                break;
+                            }
+                            if (ggml_backend_meta_get_split_state(next->src[s], false).axis != GGML_BACKEND_SPLIT_AXIS_MIRRORED) {
+                                safe = false;
+                                break;
+                            }
+                        }
+                        if (!safe) {
+                            break;
+                        }
+                        id++;
+                    }
+                };
+
+                skip_unrelated();
+                if (id + 1 >= cgraph->n_nodes) {
+                    return idr;
+                }
+                {
+                    ggml_tensor * next = cgraph->nodes[id+1];
+                    if (next->op == GGML_OP_ADD_ID && next->src[0] == node &&
+                            ggml_backend_meta_get_split_state(next->src[1], false).axis == GGML_BACKEND_SPLIT_AXIS_PARTIAL &&
+                            ggml_backend_meta_get_split_state(next->src[2], false).axis == GGML_BACKEND_SPLIT_AXIS_MIRRORED) {
+                        node = next;
+                        id++;
+                        idr = id;
+                        n_used = ggml_node_get_use_count(cgraph, id);
+                    }
+                }
+                // Chain of MULs with MIRRORED src[1]
+                while (true) {
+                    skip_unrelated();
+                    if (id + 1 >= cgraph->n_nodes) {
+                        return idr;
+                    }
+                    ggml_tensor * next = cgraph->nodes[id+1];
+                    if (next->op == GGML_OP_MUL && next->src[0] == node &&
+                            ggml_backend_meta_get_split_state(next->src[1], false).axis == GGML_BACKEND_SPLIT_AXIS_MIRRORED) {
+                        node = next;
+                        id++;
+                        idr = id;
+                        n_used = ggml_node_get_use_count(cgraph, id);
+                    } else {
+                        break;
+                    }
+                }
+
+                if (n_used != node->ne[1] || id + 2*n_used-1 >= cgraph->n_nodes) {
+                    return idr;
+                }
+                for (int32_t k = 0; k < n_used; k++) {
+                    ggml_tensor * next = cgraph->nodes[id+1];
+                    if (next->op != GGML_OP_VIEW || next->view_src != node || next->view_offs != k*node->nb[1] ||
+                            next->ne[0] != node->ne[0] || next->ne[1] != node->ne[2] || next->nb[1] != node->nb[2] ||
+                            ggml_node_get_use_count(cgraph, id+1) != 1) {
+                        return idr;
+                    }
                     id++;
-                    idr = id;
-                    n_used = ggml_node_get_use_count(cgraph, id);
                 }
-            }
-            if (id + 1 >= cgraph->n_nodes) {
-                return idr;
-            }
-            {
-                ggml_tensor * next = cgraph->nodes[id+1];
-                if (next->op == GGML_OP_MUL && next->src[0] == node &&
-                        ggml_backend_meta_get_split_state(next->src[1], false).axis == GGML_BACKEND_SPLIT_AXIS_MIRRORED) {
-                    node = next;
+                {
+                    ggml_tensor * next = cgraph->nodes[id+1];
+                    if (next->op != GGML_OP_ADD || next->src[0] != cgraph->nodes[id - (n_used-1)] ||
+                            next->src[1] != cgraph->nodes[id - (n_used-2)] || ggml_node_get_use_count(cgraph, id+1) != 1) {
+                        return idr;
+                    }
                     id++;
-                    idr = id;
-                    n_used = ggml_node_get_use_count(cgraph, id);
                 }
-            }
-
-            if (n_used != node->ne[1] || id + 2*n_used-1 >= cgraph->n_nodes) {
+                for (int32_t k = 0; k < n_used - 2; k++) {
+                    ggml_tensor * next = cgraph->nodes[id+1];
+                    if (next->op != GGML_OP_ADD || next->src[0] != cgraph->nodes[id] ||
+                            next->src[1] != cgraph->nodes[id - (n_used-2)] || ggml_node_get_use_count(cgraph, id+1) != 1) {
+                        return idr;
+                    }
+                    id++;
+                }
+                idr = id;
                 return idr;
-            }
-            for (int32_t k = 0; k < n_used; k++) {
-                ggml_tensor * next = cgraph->nodes[id+1];
-                if (next->op != GGML_OP_VIEW || next->view_src != node || next->view_offs != k*node->nb[1] ||
-                        next->ne[0] != node->ne[0] || next->ne[1] != node->ne[2] || next->nb[1] != node->nb[2] ||
-                        ggml_node_get_use_count(cgraph, id+1) != 1) {
-                    return idr;
-                }
-                id++;
-            }
-            {
-                ggml_tensor * next = cgraph->nodes[id+1];
-                if (next->op != GGML_OP_ADD || next->src[0] != cgraph->nodes[id - (n_used-1)] ||
-                        next->src[1] != cgraph->nodes[id - (n_used-2)] || ggml_node_get_use_count(cgraph, id+1) != 1) {
-                    return idr;
-                }
-                id++;
-            }
-            for (int32_t k = 0; k < n_used - 2; k++) {
-                ggml_tensor * next = cgraph->nodes[id+1];
-                if (next->op != GGML_OP_ADD || next->src[0] != cgraph->nodes[id] ||
-                        next->src[1] != cgraph->nodes[id - (n_used-2)] || ggml_node_get_use_count(cgraph, id+1) != 1) {
-                    return idr;
-                }
-                id++;
-            }
-            idr = id;
-            return idr;
-        };
+            };
 
-        int i_start = 0;
-        for (int i = 0; i < cgraph->n_nodes; i++) {
-            ggml_tensor * node = cgraph->nodes[i];
-            if (node->view_src != nullptr && node->view_src->op == GGML_OP_NONE && ggml_backend_buffer_is_host(node->view_src->buffer)) {
-                continue;
-            }
-            const ggml_backend_meta_split_state split_state = ggml_backend_meta_get_split_state(node, /*assume_sync =*/ false);
-            if (split_state.axis == GGML_BACKEND_SPLIT_AXIS_PARTIAL) {
-                max_tmp_size = std::max(max_tmp_size, ggml_nbytes(node));
-            }
-            const bool new_subgraph = i + 1 == cgraph->n_nodes || split_state.axis == GGML_BACKEND_SPLIT_AXIS_PARTIAL;
-            if (!new_subgraph) {
-                continue;
-            }
+            int i_start = 0;
+            for (int i = 0; i < cgraph->n_nodes; i++) {
+                ggml_tensor * node = cgraph->nodes[i];
+                if (node->view_src != nullptr && node->view_src->op == GGML_OP_NONE && ggml_backend_buffer_is_host(node->view_src->buffer)) {
+                    continue;
+                }
+                const ggml_backend_meta_split_state split_state = ggml_backend_meta_get_split_state(node, /*assume_sync =*/ false);
+                if (split_state.axis == GGML_BACKEND_SPLIT_AXIS_PARTIAL) {
+                    max_tmp_size = std::max(max_tmp_size, ggml_nbytes(node));
+                }
+                const bool new_subgraph = i + 1 == cgraph->n_nodes || split_state.axis == GGML_BACKEND_SPLIT_AXIS_PARTIAL;
+                if (!new_subgraph) {
+                    continue;
+                }
 
-            i = get_i_delayed(i);
+                const int i_delayed = get_i_delayed(i);
 
+                // If we can delay the AllReduce we need to consider the interaction with zero-sized tensor slices.
+                // A backend with such a slice would normally have valid data after participating in the AllReduce with a node that has
+                //     its compute flag disabled and thus gets its data zeroed out.
+                // If the AllReduce is delayed then the nodes until that point also need to have their compute flag disabled.
+                if (i_delayed > i) {
+                    for (size_t j = 0; j < n_backends; j++) {
+                        auto & bcj = backend_ctx->backend_configs[j];
+                        if ((bcj.nodes[i]->flags & GGML_TENSOR_FLAG_COMPUTE) == 0) {
+                            for (int ii = i + 1; ii <= i_delayed; ii++) {
+                                bcj.nodes[ii]->flags &= ~GGML_TENSOR_FLAG_COMPUTE;
+                            }
+                        }
+                    }
+                }
+
+                i = i_delayed;
+
+                for (size_t j = 0; j < n_backends; j++) {
+                    auto & bcj = backend_ctx->backend_configs[j];
+                    bcj.cgraphs[n_subgraphs].offset = i_start;
+                }
+                n_subgraphs++;
+                i_start = i + 1;
+            }
+            GGML_ASSERT(i_start == cgraph->n_nodes);
+        }
+
+        backend_ctx->uid         = cgraph->uid;
+        backend_ctx->n_subgraphs = n_subgraphs;
+
+        if (max_tmp_size > backend_ctx->max_tmp_size) {
             for (size_t j = 0; j < n_backends; j++) {
                 auto & bcj = backend_ctx->backend_configs[j];
-                bcj.cgraphs[n_subgraphs].offset = i_start;
+                for (size_t i = 0; i < backend_ctx->n_reduce_steps; i++) {
+                    bcj.bufs[i].reset(ggml_backend_alloc_buffer(bcj.backend, max_tmp_size));
+                }
             }
-            n_subgraphs++;
-            i_start = i + 1;
+            backend_ctx->max_tmp_size = max_tmp_size;
         }
-        GGML_ASSERT(i_start == cgraph->n_nodes);
-    }
 
-    if (max_tmp_size > backend_ctx->max_tmp_size) {
-        for (size_t j = 0; j < n_backends; j++) {
-            auto & bcj = backend_ctx->backend_configs[j];
-            bcj.buf.reset(ggml_backend_alloc_buffer(bcj.backend, max_tmp_size));
-        }
-        backend_ctx->max_tmp_size = max_tmp_size;
-    }
-
-
-    if (max_nnodes_raised || n_subgraphs > backend_ctx->max_subgraphs) {
-        backend_ctx->max_subgraphs = std::max(backend_ctx->max_subgraphs, n_subgraphs);
-        const size_t n_reduce_steps = backend_ctx->n_reduce_steps();
-        const size_t n_nodes_per_device = 2 * n_reduce_steps; // tmp + ADD per step
-        const size_t n_cgraphs_per_device = n_reduce_steps;    // 1 ADD graph per step
-        const size_t mem_per_device_graphs_main = backend_ctx->max_subgraphs*ggml_graph_overhead_custom(backend_ctx->max_nnodes, cgraph->grads);
-        const size_t mem_per_device_graphs_aux = n_cgraphs_per_device*backend_ctx->max_subgraphs*ggml_graph_overhead_custom(1, cgraph->grads);
-        const size_t mem_per_device_nodes_aux = n_nodes_per_device*backend_ctx->max_subgraphs*ggml_tensor_overhead();
-        ggml_init_params params = {
-            /*.mem_size   =*/ n_backends * (mem_per_device_graphs_main + mem_per_device_graphs_aux + mem_per_device_nodes_aux),
-            /*.mem_buffer =*/ nullptr,
-            /*.no_alloc   =*/ true,
-        };
-        backend_ctx->ctx.reset(ggml_init(params));
-        for (size_t j = 0; j < n_backends; j++) {
-            auto & bcj = backend_ctx->backend_configs[j];
-            for (size_t i = 0; i < n_subgraphs; i++) {
-                bcj.cgraphs[i].cgraph_main = ggml_new_graph_custom(backend_ctx->ctx.get(), cgraph->n_nodes, /*grads =*/ false);
+        if (max_nnodes_raised || n_subgraphs > backend_ctx->max_subgraphs) {
+            backend_ctx->max_subgraphs = std::max(backend_ctx->max_subgraphs, n_subgraphs);
+            const size_t n_nodes_per_device = 3 * backend_ctx->n_reduce_steps; // tmp + ADD (+zeroing) graph per step and device
+            const size_t n_cgraphs_per_device = 2 * backend_ctx->n_reduce_steps; // ADD ( + zeroing) graph per step and device
+            const size_t mem_per_device_graphs_main = backend_ctx->max_subgraphs*ggml_graph_overhead_custom(backend_ctx->max_nnodes, cgraph->grads);
+            const size_t mem_per_device_graphs_aux = n_cgraphs_per_device*backend_ctx->max_subgraphs*ggml_graph_overhead_custom(1, cgraph->grads);
+            const size_t mem_per_device_nodes_aux = n_nodes_per_device*backend_ctx->max_subgraphs*ggml_tensor_overhead();
+            ggml_init_params params = {
+                /*.mem_size   =*/ n_backends * (mem_per_device_graphs_main + mem_per_device_graphs_aux + mem_per_device_nodes_aux),
+                /*.mem_buffer =*/ nullptr,
+                /*.no_alloc   =*/ true,
+            };
+            backend_ctx->ctx.reset(ggml_init(params));
+            for (size_t j = 0; j < n_backends; j++) {
+                auto & bcj = backend_ctx->backend_configs[j];
+                for (size_t i = 0; i < n_subgraphs; i++) {
+                    bcj.cgraphs[i].cgraph_main = ggml_new_graph_custom(backend_ctx->ctx.get(), cgraph->n_nodes, /*grads =*/ false);
+                }
+            }
+            backend_ctx->cgraphs_aux.resize(n_backends*n_cgraphs_per_device*backend_ctx->max_subgraphs);
+            for (size_t k = 0; k < backend_ctx->cgraphs_aux.size(); k++) {
+                backend_ctx->cgraphs_aux[k] = ggml_new_graph_custom(backend_ctx->ctx.get(), 1, cgraph->grads);
+            }
+            backend_ctx->nodes_aux.resize(n_backends*n_nodes_per_device*backend_ctx->max_subgraphs);
+            for (size_t k = 0; k < backend_ctx->nodes_aux.size(); k++) {
+                backend_ctx->nodes_aux[k] = ggml_new_tensor_1d(backend_ctx->ctx.get(), GGML_TYPE_F32, 1);
             }
         }
-        backend_ctx->cgraphs_aux.resize(n_backends*n_cgraphs_per_device*backend_ctx->max_subgraphs);
-        for (size_t k = 0; k < backend_ctx->cgraphs_aux.size(); k++) {
-            backend_ctx->cgraphs_aux[k] = ggml_new_graph_custom(backend_ctx->ctx.get(), 1, cgraph->grads);
-        }
-        backend_ctx->nodes_aux.resize(n_backends*n_nodes_per_device*backend_ctx->max_subgraphs);
-        for (size_t k = 0; k < backend_ctx->nodes_aux.size(); k++) {
-            backend_ctx->nodes_aux[k] = ggml_new_tensor_1d(backend_ctx->ctx.get(), GGML_TYPE_F32, 1);
-        }
-    }
 
-    for (size_t j = 0; j < n_backends; j++) {
-        auto & bcj = backend_ctx->backend_configs[j];
-        for (size_t i_graph = 0; i_graph < n_subgraphs; i_graph++) {
-            ggml_cgraph * cgraph_ij = bcj.cgraphs[i_graph].cgraph_main;
-            const size_t i_node_start = bcj.cgraphs[i_graph].offset;
-            const size_t i_node_stop = i_graph + 1 < n_subgraphs ? bcj.cgraphs[i_graph + 1].offset : cgraph->n_nodes;
-            cgraph_ij->n_nodes = i_node_stop - i_node_start;
-            ggml_hash_set_reset(&cgraph_ij->visited_hash_set);
-            for (size_t i_node = i_node_start; i_node < i_node_stop; i_node++) {
-                ggml_tensor * node_ij = bcj.nodes[i_node];
-                cgraph_ij->nodes[i_node - i_node_start] = node_ij;
-                const size_t hash_pos_orig = ggml_hash_find(&cgraph->visited_hash_set, cgraph->nodes[i_node]);
-                const size_t hash_pos_ij = ggml_hash_insert(&cgraph_ij->visited_hash_set, node_ij);
-                cgraph_ij->use_counts[hash_pos_ij] = cgraph->use_counts[hash_pos_orig];
+        for (size_t j = 0; j < n_backends; j++) {
+            auto & bcj = backend_ctx->backend_configs[j];
+            for (size_t i_graph = 0; i_graph < n_subgraphs; i_graph++) {
+                ggml_cgraph * cgraph_ij = bcj.cgraphs[i_graph].cgraph_main;
+                const size_t i_node_start = bcj.cgraphs[i_graph].offset;
+                const size_t i_node_stop = i_graph + 1 < n_subgraphs ? bcj.cgraphs[i_graph + 1].offset : cgraph->n_nodes;
+                cgraph_ij->n_nodes = i_node_stop - i_node_start;
+                ggml_hash_set_reset(&cgraph_ij->visited_hash_set);
+                for (size_t i_node = i_node_start; i_node < i_node_stop; i_node++) {
+                    ggml_tensor * node_ij = bcj.nodes[i_node];
+                    cgraph_ij->nodes[i_node - i_node_start] = node_ij;
+                    const size_t hash_pos_orig = ggml_hash_find(&cgraph->visited_hash_set, cgraph->nodes[i_node]);
+                    const size_t hash_pos_ij = ggml_hash_insert(&cgraph_ij->visited_hash_set, node_ij);
+                    cgraph_ij->use_counts[hash_pos_ij] = cgraph->use_counts[hash_pos_orig];
+                }
+                cgraph_ij->uid = ggml_graph_next_uid();
             }
         }
     }
@@ -1761,11 +1920,6 @@ static enum ggml_status ggml_backend_meta_graph_compute(ggml_backend_t backend,
     size_t iga = 0; // i graph aux
     size_t ina = 0; // i node aux
 
-    // FIXME usage_counts
-    auto get_cgraph_aux = [&]() -> ggml_cgraph * {
-        ggml_cgraph * ret = backend_ctx->cgraphs_aux[iga++];
-        return ret;
-    };
     auto get_node_aux = [&](ggml_tensor * t) -> ggml_tensor * {
         ggml_tensor * ret = backend_ctx->nodes_aux[ina++];
         memset(ret, 0, sizeof(ggml_tensor));
@@ -1777,75 +1931,110 @@ static enum ggml_status ggml_backend_meta_graph_compute(ggml_backend_t backend,
         }
         return ret;
     };
+    auto set_tmp_data = [&](ggml_tensor * tensor, const size_t j, const size_t i_buf) {
+        auto & bcj = backend_ctx->backend_configs[j];
+        ggml_backend_buffer_ptr & buf_ptr = bcj.bufs[i_buf];
+        if (!buf_ptr || ggml_backend_buffer_get_size(buf_ptr.get()) < backend_ctx->max_tmp_size) {
+            buf_ptr.reset(ggml_backend_alloc_buffer(bcj.backend, backend_ctx->max_tmp_size));
+        }
+        tensor->buffer = buf_ptr.get();
+        tensor->data   = ggml_backend_buffer_get_base(buf_ptr.get());
+    };
+    // FIXME usage_counts
+    auto get_cgraph_aux = [&]() -> ggml_cgraph * {
+        ggml_cgraph * ret = backend_ctx->cgraphs_aux[iga++];
+        return ret;
+    };
 
     // Preferentially use backend-specific allreduce_tensor_async (e.g. NCCL for CUDA), use a generic fallback if unavailable:
     auto allreduce_fallback = [&](size_t i) -> ggml_status {
         std::vector<ggml_cgraph *> step_cgraphs(n_backends, nullptr);
 
-        for (size_t offset_j = 1; offset_j < n_backends; offset_j *= 2) {
+        // Zero out nodes that were disabled due to having a zero-sized slice:
+        for (size_t j = 0; j < n_backends; j++) {
+            auto & bcj = backend_ctx->backend_configs[j];
+            ggml_tensor * node = bcj.cgraphs[i].cgraph_main->nodes[bcj.cgraphs[i].cgraph_main->n_nodes - 1];
+            if (node->flags & GGML_TENSOR_FLAG_COMPUTE) {
+                continue;
+            }
+            ggml_tensor * node_zero = get_node_aux(node);
+            node_zero->op = GGML_OP_SCALE; // FIXME 0.0f * NaN == NaN
+            node_zero->src[0] = node;
+            ggml_set_op_params_f32(node_zero, 0, 0.0f);
+            node_zero->data = node->data;
+            node_zero->flags |= GGML_TENSOR_FLAG_COMPUTE;
+
+            step_cgraphs[j] = get_cgraph_aux();
+            step_cgraphs[j]->nodes[0] = node_zero;
+            step_cgraphs[j]->n_nodes = 1;
+            const ggml_status status = ggml_backend_graph_compute_async(bcj.backend, step_cgraphs[j]);
+            if (status != GGML_STATUS_SUCCESS) {
+                return status;
+            }
+        }
+        std::fill(step_cgraphs.begin(), step_cgraphs.end(), nullptr);
+
+        auto push_data = [&](const size_t j_src, const size_t j_dst, const size_t i_buf) {
+            assert(step_cgraphs[j_dst] == nullptr);
+            auto & bcj_src = backend_ctx->backend_configs[j_src];
+            auto & bcj_dst = backend_ctx->backend_configs[j_dst];
+
+            ggml_tensor * node_src = bcj_src.cgraphs[i].cgraph_main->nodes[bcj_src.cgraphs[i].cgraph_main->n_nodes - 1];
+            ggml_tensor * node_dst = bcj_dst.cgraphs[i].cgraph_main->nodes[bcj_dst.cgraphs[i].cgraph_main->n_nodes - 1];
+            GGML_ASSERT(ggml_is_contiguous(node_src));
+            GGML_ASSERT(ggml_is_contiguous(node_dst));
+
+            ggml_tensor * node_tmp = get_node_aux(node_dst);
+            set_tmp_data(node_tmp, j_dst, i_buf);
+
+            ggml_backend_tensor_copy_async(bcj_src.backend, bcj_dst.backend, node_src, node_tmp);
+
+            ggml_tensor * node_red = get_node_aux(node_dst);
+            node_red->view_src = node_dst->view_src == nullptr ? node_dst : node_dst->view_src;
+            node_red->view_offs = node_dst->view_offs;
+            node_red->op = GGML_OP_ADD;
+            node_red->src[0] = node_dst;
+            node_red->src[1] = node_tmp;
+            node_red->flags |= GGML_TENSOR_FLAG_COMPUTE;
+            ggml_backend_view_init(node_red);
+
+            ggml_cgraph * cgraph_aux = get_cgraph_aux();
+            cgraph_aux->nodes[0] = node_red;
+            cgraph_aux->n_nodes = 1;
+            step_cgraphs[j_dst] = cgraph_aux;
+        };
+
+        size_t offset_j = n_backends/2;
+        while ((offset_j & (offset_j - 1)) != 0) {
+            offset_j--;
+        }
+        const size_t offset_j_max = offset_j;
+        size_t i_buf = 0;
+
+        // If n_backends is not a power of 2, fold in the excess prior to butterfly reduction:
+        for (size_t j_src = 2*offset_j_max; j_src < n_backends; j_src++) {
+            const size_t j_dst = j_src - 2*offset_j_max;
+            push_data(j_src, j_dst, i_buf);
+            const ggml_status status = ggml_backend_graph_compute_async(backend_ctx->backend_configs[j_dst].backend, step_cgraphs[j_dst]);
+            if (status != GGML_STATUS_SUCCESS) {
+                return status;
+            }
+            i_buf = 1;
+        }
+
+        // Butterfly reduction:
+        for (; offset_j >= 1; offset_j /= 2) {
             std::fill(step_cgraphs.begin(), step_cgraphs.end(), nullptr);
 
-            for (size_t j = 0; j < n_backends; j++) {
+            for (size_t j = 0; j < 2*offset_j_max; j++) {
                 const size_t j_other = j ^ offset_j;
-                if (j_other > j) {
+                if (j_other >= n_backends) {
                     continue;
                 }
-
-                auto & bcj1 = backend_ctx->backend_configs[j];
-                auto & bcj2 = backend_ctx->backend_configs[j_other];
-
-                ggml_tensor * node1 = bcj1.cgraphs[i].cgraph_main->nodes[bcj1.cgraphs[i].cgraph_main->n_nodes - 1];
-                ggml_tensor * node2 = bcj2.cgraphs[i].cgraph_main->nodes[bcj2.cgraphs[i].cgraph_main->n_nodes - 1];
-                GGML_ASSERT(ggml_is_contiguous(node1));
-                GGML_ASSERT(ggml_is_contiguous(node2));
-
-                // Tmp tensors to receive P2P copies
-                ggml_tensor * node_tmp_1 = get_node_aux(node1);
-                node_tmp_1->buffer = bcj1.buf.get();
-                node_tmp_1->data = ggml_backend_buffer_get_base(bcj1.buf.get());
-
-                ggml_tensor * node_tmp_2 = get_node_aux(node2);
-                node_tmp_2->buffer = bcj2.buf.get();
-                node_tmp_2->data = ggml_backend_buffer_get_base(bcj2.buf.get());
-
-                // 2 P2P copies: exchange full buffers
-                ggml_backend_tensor_copy_async(bcj1.backend, bcj2.backend, node1, node_tmp_2);
-                ggml_backend_tensor_copy_async(bcj2.backend, bcj1.backend, node2, node_tmp_1);
-
-                // Local ADD: node1 += tmp1 (in-place via view)
-                ggml_tensor * node_red_1 = get_node_aux(node1);
-                node_red_1->view_src = node1->view_src == nullptr ? node1 : node1->view_src;
-                node_red_1->view_offs = node1->view_offs;
-                node_red_1->op = GGML_OP_ADD;
-                node_red_1->src[0] = node1;
-                node_red_1->src[1] = node_tmp_1;
-                node_red_1->flags |= GGML_TENSOR_FLAG_COMPUTE;
-                ggml_backend_view_init(node_red_1);
-
-                // Local ADD: node2 += tmp2 (in-place via view)
-                ggml_tensor * node_red_2 = get_node_aux(node2);
-                node_red_2->view_src = node2->view_src == nullptr ? node2 : node2->view_src;
-                node_red_2->view_offs = node2->view_offs;
-                node_red_2->op = GGML_OP_ADD;
-                node_red_2->src[0] = node2;
-                node_red_2->src[1] = node_tmp_2;
-                node_red_2->flags |= GGML_TENSOR_FLAG_COMPUTE;
-                ggml_backend_view_init(node_red_2);
-
-                // Build 1-node cgraphs for the ADD ops
-                ggml_cgraph * cgraph_aux_1 = get_cgraph_aux();
-                cgraph_aux_1->nodes[0] = node_red_1;
-                cgraph_aux_1->n_nodes = 1;
-                step_cgraphs[j] = cgraph_aux_1;
-
-                ggml_cgraph * cgraph_aux_2 = get_cgraph_aux();
-                cgraph_aux_2->nodes[0] = node_red_2;
-                cgraph_aux_2->n_nodes = 1;
-                step_cgraphs[j_other] = cgraph_aux_2;
+                push_data(j, j_other, i_buf);
             }
 
-            // Execute local ADDs for this step
-            for (size_t j = 0; j < n_backends; j++) {
+            for (size_t j = 0; j < 2*offset_j_max; j++) {
                 if (step_cgraphs[j] == nullptr) {
                     continue;
                 }
@@ -1855,12 +2044,25 @@ static enum ggml_status ggml_backend_meta_graph_compute(ggml_backend_t backend,
                     return status;
                 }
             }
+            i_buf++;
         }
+        assert(i_buf == backend_ctx->n_reduce_steps);
+
+        // If n_backends is not a power of 2, copy back the reduced tensors to the excess:
+        for (size_t j = 2*offset_j_max; j < n_backends; j++) {
+            auto & bcj_src = backend_ctx->backend_configs[j - 2*offset_j_max];
+            auto & bcj_dst = backend_ctx->backend_configs[j];
+
+            ggml_tensor * node_src = bcj_src.cgraphs[i].cgraph_main->nodes[bcj_src.cgraphs[i].cgraph_main->n_nodes - 1];
+            ggml_tensor * node_dst = bcj_dst.cgraphs[i].cgraph_main->nodes[bcj_dst.cgraphs[i].cgraph_main->n_nodes - 1];
+            ggml_backend_tensor_copy_async(bcj_src.backend, bcj_dst.backend, node_src, node_dst);
+        }
+
         return GGML_STATUS_SUCCESS;
     };
 
 
-    for (size_t i = 0; i < n_subgraphs; i++) {
+    for (size_t i = 0; i < backend_ctx->n_subgraphs; i++) {
         for (size_t j = 0; j < n_backends; j++) {
             auto & bcj = backend_ctx->backend_configs[j];
             const ggml_status status = ggml_backend_graph_compute_async(bcj.backend, bcj.cgraphs[i].cgraph_main);
@@ -1869,7 +2071,7 @@ static enum ggml_status ggml_backend_meta_graph_compute(ggml_backend_t backend,
             }
         }
 
-        if (n_backends > 1 && i < n_subgraphs - 1) {
+        if (n_backends > 1 && i < backend_ctx->n_subgraphs - 1) {
             bool backend_allreduce_success = false;
             if (backend_ctx->comm_ctx) {
                 std::vector<ggml_tensor *> nodes;
@@ -1898,8 +2100,8 @@ static const ggml_backend_i ggml_backend_meta_i = {
     /* .free                    = */ ggml_backend_meta_free,
     /* .set_tensor_async        = */ ggml_backend_meta_set_tensor_async,
     /* .get_tensor_async        = */ ggml_backend_meta_get_tensor_async,
-    /* .get_tensor_2d_async     = */ nullptr,
     /* .set_tensor_2d_async     = */ nullptr,
+    /* .get_tensor_2d_async     = */ nullptr,
     /* .cpy_tensor_async        = */ nullptr,
     /* .synchronize             = */ ggml_backend_meta_synchronize,
     /* .graph_plan_create       = */ nullptr,

ggml/src/ggml-backend-reg.cpp

@@ -181,6 +181,12 @@ struct ggml_backend_registry {
             return;
         }
 
+        for (auto & entry : backends) {
+            if (entry.reg == reg) {
+                return;
+            }
+        }
+
 #ifndef NDEBUG
         GGML_LOG_DEBUG("%s: registered backend %s (%zu devices)\n",
             __func__, ggml_backend_reg_name(reg), ggml_backend_reg_dev_count(reg));
@@ -192,6 +198,12 @@ struct ggml_backend_registry {
     }
 
     void register_device(ggml_backend_dev_t device) {
+        for (auto & dev : devices) {
+            if (dev == device) {
+                return;
+            }
+        }
+
 #ifndef NDEBUG
         GGML_LOG_DEBUG("%s: registered device %s (%s)\n", __func__, ggml_backend_dev_name(device), ggml_backend_dev_description(device));
 #endif

ggml/src/ggml-backend.cpp

@@ -1030,6 +1030,8 @@ void ggml_backend_sched_split_graph(ggml_backend_sched_t sched, struct ggml_cgra
         GGML_ABORT("%s: failed to initialize context\n", __func__);
     }
 
+    graph->uid = ggml_graph_next_uid();
+
     // pass 1: assign backends to ops with pre-allocated inputs
     for (int i = 0; i < graph->n_leafs; i++) {
         struct ggml_tensor * leaf = graph->leafs[i];
@@ -1477,6 +1479,11 @@ void ggml_backend_sched_split_graph(ggml_backend_sched_t sched, struct ggml_cgra
         assert(graph_copy->size > graph_copy->n_leafs);
         graph_copy->leafs[graph_copy->n_leafs++] = leaf;
     }
+
+    // set ids for all splits
+    for (int i = 0; i < sched->n_splits; ++i) {
+        sched->splits[i].graph.uid = ggml_graph_next_uid();
+    }
 }
 
 static bool ggml_backend_sched_alloc_splits(ggml_backend_sched_t sched) {

ggml/src/ggml-blas/ggml-blas.cpp

@@ -262,9 +262,9 @@ static struct ggml_backend_i blas_backend_i = {
     /* .get_name                = */ ggml_backend_blas_get_name,
     /* .free                    = */ ggml_backend_blas_free,
     /* .set_tensor_async        = */ NULL,
-    /* .get_tensor_2d_async     = */ NULL,
-    /* .set_tensor_2d_async     = */ NULL,
     /* .get_tensor_async        = */ NULL,
+    /* .set_tensor_2d_async     = */ NULL,
+    /* .get_tensor_2d_async     = */ NULL,
     /* .cpy_tensor_async        = */ NULL,
     /* .synchronize             = */ NULL,
     /* .graph_plan_create       = */ NULL,