llama : remove beam search (#7736 )

readme : remove obsolete Zig instructions (#7471 )
llama-bench : allow using a different printer for stderr with -oe (#7722 )
2026-05-19 07:24:07 +00:00 · 2024-06-04 21:23:05 +03:00 · 2024-06-04 19:43:01 +03:00 · 2024-06-04 14:32:42 +02:00 · 2024-06-04 14:09:15 +02:00 · 2024-06-04 21:21:26 +10:00
26 changed files with 73895 additions and 14583 deletions
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -294,12 +294,22 @@ jobs:

      - name: Build
        id: cmake_build
+        if: ${{ matrix.sanitizer != 'THREAD' }}
        run: |
          mkdir build
          cd build
          cmake .. -DLLAMA_FATAL_WARNINGS=ON -DLLAMA_SANITIZE_${{ matrix.sanitizer }}=ON -DCMAKE_BUILD_TYPE=${{ matrix.build_type }}
          cmake --build . --config ${{ matrix.build_type }} -j $(nproc)

+      - name: Build (no OpenMP)
+        id: cmake_build_no_openmp
+        if: ${{ matrix.sanitizer == 'THREAD' }}
+        run: |
+          mkdir build
+          cd build
+          cmake .. -DLLAMA_FATAL_WARNINGS=ON -DLLAMA_SANITIZE_${{ matrix.sanitizer }}=ON -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} -DLLAMA_OPENMP=OFF
+          cmake --build . --config ${{ matrix.build_type }} -j $(nproc)
+
      - name: Test
        id: cmake_test
        run: |
--- a/.gitignore
+++ b/.gitignore
@@ -34,9 +34,11 @@ ggml-metal-embed.metal
 lcov-report/
 gcovr-report/

+tags
 build*
 !build.zig
 cmake-build-*
+android-ndk-*
 out/
 tmp/

--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -126,6 +126,7 @@ set(LLAMA_METAL_MACOSX_VERSION_MIN "" CACHE STRING
 set(LLAMA_METAL_STD "" CACHE STRING          "llama: metal standard version (-std flag)")
 option(LLAMA_KOMPUTE                         "llama: use Kompute"                               OFF)
 option(LLAMA_RPC                             "llama: use RPC"                                   OFF)
+option(LLAMA_OPENMP                          "llama: use OpenMP"                                ON)
 option(LLAMA_SYCL                            "llama: use SYCL"                                  OFF)
 option(LLAMA_SYCL_F16                        "llama: use 16 bit floats for sycl calculations"   OFF)
 set(LLAMA_SYCL_TARGET   "INTEL" CACHE STRING "llama: sycl target device")
@@ -296,6 +297,17 @@ if (LLAMA_METAL)
        )
 endif()

+if (LLAMA_OPENMP)
+    find_package(OpenMP)
+    if (OpenMP_FOUND)
+        message(STATUS "OpenMP found")
+        add_compile_definitions(GGML_USE_OPENMP)
+        set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} OpenMP::OpenMP_C OpenMP::OpenMP_CXX)
+    else()
+        message(WARNING "OpenMP not found")
+    endif()
+endif()
+
 if (LLAMA_BLAS)
    if (LLAMA_STATIC)
        set(BLA_STATIC ON)
@@ -545,12 +557,17 @@ if (LLAMA_VULKAN)
 endif()

 if (LLAMA_HIPBLAS)
-    if ($ENV{ROCM_PATH})
-        set(ROCM_PATH $ENV{ROCM_PATH})
+    if (NOT EXISTS $ENV{ROCM_PATH})
+        if (NOT EXISTS /opt/rocm)
+            set(ROCM_PATH /usr)
+        else()
+            set(ROCM_PATH /opt/rocm)
+        endif()
    else()
-        set(ROCM_PATH /opt/rocm)
+        set(ROCM_PATH $ENV{ROCM_PATH})
    endif()
    list(APPEND CMAKE_PREFIX_PATH ${ROCM_PATH})
+    list(APPEND CMAKE_PREFIX_PATH "${ROCM_PATH}/lib64/cmake")

    # CMake on Windows doesn't support the HIP language yet
    if(WIN32)
@@ -1373,6 +1390,13 @@ if (LLAMA_METAL)
    endif()
 endif()

+configure_file(cmake/llama.pc.in
+        "${CMAKE_CURRENT_BINARY_DIR}/llama.pc"
+        @ONLY)
+
+install(FILES "${CMAKE_CURRENT_BINARY_DIR}/llama.pc"
+        DESTINATION lib/pkgconfig)
+
 #
 # programs, examples and tests
 #
--- a/56
+++ b/56
@@ -1,7 +1,7 @@
 # Define the default target now so that it is always the first target
 BUILD_TARGETS = \
 	main quantize quantize-stats perplexity imatrix embedding vdot q8dot train-text-from-scratch convert-llama2c-to-ggml \
-	simple batched batched-bench save-load-state server gguf gguf-split eval-callback llama-bench libllava.a llava-cli baby-llama beam-search  \
+	simple batched batched-bench save-load-state server gguf gguf-split eval-callback llama-bench libllava.a llava-cli baby-llama \
 	retrieval speculative infill tokenize benchmark-matmult parallel finetune export-lora lookahead lookup passkey gritlm tests/test-c.o

 # Binaries only useful for tests
@@ -57,6 +57,8 @@ ifeq ($(UNAME_S),Darwin)
 		LLAMA_METAL := 1
 	endif

+	LLAMA_NO_OPENMP := 1
+
 	ifneq ($(UNAME_P),arm)
 		SYSCTL_M := $(shell sysctl -n hw.optional.arm64 2>/dev/null)
 		ifeq ($(SYSCTL_M),1)
@@ -67,6 +69,10 @@ ifeq ($(UNAME_S),Darwin)
 	endif
 endif

+ifdef LLAMA_RPC
+	BUILD_TARGETS += rpc-server
+endif
+
 default: $(BUILD_TARGETS)

 test: $(TEST_TARGETS)
@@ -135,12 +141,16 @@ MK_NVCCFLAGS = -std=c++11
 ifdef LLAMA_FAST
 MK_CFLAGS     += -Ofast
 HOST_CXXFLAGS += -Ofast
+ifndef LLAMA_DEBUG
 MK_NVCCFLAGS  += -O3
+endif # LLAMA_DEBUG
 else
 MK_CFLAGS     += -O3
 MK_CXXFLAGS   += -O3
+ifndef LLAMA_DEBUG
 MK_NVCCFLAGS  += -O3
-endif
+endif # LLAMA_DEBUG
+endif # LLAMA_FAST

 ifndef LLAMA_NO_CCACHE
 CCACHE := $(shell which ccache)
@@ -201,9 +211,10 @@ ifdef LLAMA_SCHED_MAX_COPIES
 endif

 ifdef LLAMA_DEBUG
-	MK_CFLAGS   += -O0 -g
-	MK_CXXFLAGS += -O0 -g
-	MK_LDFLAGS  += -g
+	MK_CFLAGS    += -O0 -g
+	MK_CXXFLAGS  += -O0 -g
+	MK_LDFLAGS   += -g
+	MK_NVCCFLAGS += -O0 -g

 	ifeq ($(UNAME_S),Linux)
 		MK_CPPFLAGS += -D_GLIBCXX_ASSERTIONS
@@ -400,6 +411,12 @@ ifndef LLAMA_NO_ACCELERATE
 	endif
 endif # LLAMA_NO_ACCELERATE

+ifndef LLAMA_NO_OPENMP
+	MK_CPPFLAGS += -DGGML_USE_OPENMP
+	MK_CFLAGS   += -fopenmp
+	MK_CXXFLAGS += -fopenmp
+endif # LLAMA_NO_OPENMP
+
 ifdef LLAMA_OPENBLAS
 	MK_CPPFLAGS += -DGGML_USE_OPENBLAS $(shell pkg-config --cflags-only-I openblas)
 	MK_CFLAGS   += $(shell pkg-config --cflags-only-other openblas)
@@ -416,6 +433,11 @@ ifdef LLAMA_BLIS
 	MK_LDFLAGS  += -lblis -L/usr/local/lib
 endif # LLAMA_BLIS

+ifdef LLAMA_RPC
+	MK_CPPFLAGS   += -DGGML_USE_RPC
+	OBJS          += ggml-rpc.o
+endif # LLAMA_RPC
+
 ifdef LLAMA_CUBLAS
 # LLAMA_CUBLAS is deprecated and will be removed in the future
 	LLAMA_CUDA := 1
@@ -641,11 +663,26 @@ ggml-metal-embed.o: ggml-metal.metal ggml-common.h
 endif
 endif # LLAMA_METAL

+OBJS += ggml-alloc.o ggml-backend.o ggml-quants.o unicode.o unicode-data.o
+COMMON_H_DEPS = common/common.h common/sampling.h common/log.h llama.h
+COMMON_DEPS   = common.o sampling.o grammar-parser.o build-info.o json-schema-to-grammar.o
+
 ifndef LLAMA_NO_LLAMAFILE
 sgemm.o: sgemm.cpp sgemm.h ggml.h
 	$(CXX) $(CXXFLAGS) -c $< -o $@
 endif

+ifdef LLAMA_RPC
+ggml-rpc.o: ggml-rpc.cpp ggml-rpc.h
+	$(CXX) $(CXXFLAGS) -c $< -o $@
+
+rpc-server.o: examples/rpc/rpc-server.cpp ggml-rpc.h
+	$(CXX) $(CXXFLAGS) -c $< -o $@
+
+rpc-server: rpc-server.o ggml.o llama.o $(COMMON_DEPS) $(OBJS)
+	$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
+endif # LLAMA_RPC
+
 GF_CC := $(CC)
 include scripts/get-flags.mk

@@ -725,14 +762,9 @@ unicode.o: unicode.cpp unicode.h
 unicode-data.o: unicode-data.cpp unicode-data.h
 	$(CXX) $(CXXFLAGS) -c $< -o $@

-OBJS += ggml-alloc.o ggml-backend.o ggml-quants.o unicode.o unicode-data.o
-
 llama.o: llama.cpp unicode.h ggml.h ggml-alloc.h ggml-backend.h ggml-cuda.h ggml-metal.h llama.h
 	$(CXX) $(CXXFLAGS) -c $< -o $@

-COMMON_H_DEPS = common/common.h common/sampling.h common/log.h llama.h
-COMMON_DEPS   = common.o sampling.o grammar-parser.o build-info.o json-schema-to-grammar.o
-
 common.o: common/common.cpp $(COMMON_H_DEPS)
 	$(CXX) $(CXXFLAGS) -c $< -o $@

@@ -882,10 +914,6 @@ baby-llama: examples/baby-llama/baby-llama.cpp ggml.o llama.o $(COMMON_DEPS) tra
 	$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
 	$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)

-beam-search: examples/beam-search/beam-search.cpp ggml.o llama.o $(COMMON_DEPS) $(OBJS)
-	$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
-	$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
-
 finetune: examples/finetune/finetune.cpp ggml.o llama.o $(COMMON_DEPS) train.o $(OBJS)
 	$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
 	$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
--- a/README.md
+++ b/README.md
@@ -364,17 +364,6 @@ In order to build llama.cpp you have four different options.
      cmake --build build --config Debug
      ```

- Using `Zig` (version 0.11 or later):
-
-    Building for optimization levels and CPU features can be accomplished using standard build arguments, for example AVX2, FMA, F16C,
-    it's also possible to cross compile for other operating systems and architectures:
-
-    ```bash
-    zig build -Doptimize=ReleaseFast -Dtarget=x86_64-windows-gnu -Dcpu=x86_64+avx2+fma+f16c
-    ```
-
-    The `zig targets` command will give you valid options to use.
-
 -   Using `gmake` (FreeBSD):

    1. Install and activate [DRM in FreeBSD](https://wiki.freebsd.org/Graphics)
--- a/cmake/arm64-windows-llvm.cmake
+++ b/cmake/arm64-windows-llvm.cmake
@@ -9,7 +9,7 @@ set( CMAKE_CXX_COMPILER  clang++ )
 set( CMAKE_C_COMPILER_TARGET   ${target} )
 set( CMAKE_CXX_COMPILER_TARGET ${target} )

-set( arch_c_flags "-march=armv8.7-a -fvectorize -ffp-model=fast" )
+set( arch_c_flags "-march=armv8.7-a -fvectorize -ffp-model=fast -fno-finite-math-only" )
 set( warn_c_flags "-Wno-format -Wno-unused-variable -Wno-unused-function -Wno-gnu-zero-variadic-macro-arguments" )

 set( CMAKE_C_FLAGS_INIT   "${arch_c_flags} ${warn_c_flags}" )
--- a/cmake/llama.pc.in
+++ b/cmake/llama.pc.in
@@ -0,0 +1,10 @@
+prefix=@CMAKE_INSTALL_PREFIX@
+exec_prefix=${prefix}
+libdir=${exec_prefix}/lib
+includedir=${prefix}/include
+
+Name: llama
+Description: Port of Facebook's LLaMA model in C/C++
+Version: @PROJECT_VERSION@
+Libs: -L${libdir} -lllama
+Cflags: -I${includedir}
--- a/common/common.cpp
+++ b/common/common.cpp
@@ -1002,9 +1002,9 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
            return true;
        }
        params.main_gpu = std::stoi(argv[i]);
-#ifndef GGML_USE_CUDA_SYCL
-        fprintf(stderr, "warning: llama.cpp was compiled without CUDA/SYCL. Setting the main GPU has no effect.\n");
-#endif // GGML_USE_CUDA_SYCL
+#ifndef GGML_USE_CUDA_SYCL_VULKAN
+        fprintf(stderr, "warning: llama.cpp was compiled without CUDA/SYCL/Vulkan. Setting the main GPU has no effect.\n");
+#endif // GGML_USE_CUDA_SYCL_VULKAN
        return true;
    }
    if (arg == "--split-mode" || arg == "-sm") {
@@ -1030,9 +1030,9 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
            invalid_param = true;
            return true;
        }
-#ifndef GGML_USE_CUDA_SYCL
-        fprintf(stderr, "warning: llama.cpp was compiled without CUDA/SYCL. Setting the split mode has no effect.\n");
-#endif // GGML_USE_CUDA_SYCL
+#ifndef GGML_USE_CUDA_SYCL_VULKAN
+        fprintf(stderr, "warning: llama.cpp was compiled without CUDA/SYCL/Vulkan. Setting the split mode has no effect.\n");
+#endif // GGML_USE_CUDA_SYCL_VULKAN
        return true;
    }
    if (arg == "--tensor-split" || arg == "-ts") {
--- a/examples/CMakeLists.txt
+++ b/examples/CMakeLists.txt
@@ -15,7 +15,6 @@ else()
    add_subdirectory(baby-llama)
    add_subdirectory(batched)
    add_subdirectory(batched-bench)
-    add_subdirectory(beam-search)
    add_subdirectory(benchmark)
    add_subdirectory(convert-llama2c-to-ggml)
    add_subdirectory(embedding)
--- a/examples/beam-search/CMakeLists.txt
+++ b/examples/beam-search/CMakeLists.txt
@@ -1,5 +0,0 @@
-set(TARGET beam-search)
-add_executable(${TARGET} beam-search.cpp)
-install(TARGETS ${TARGET} RUNTIME)
-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
-target_compile_features(${TARGET} PRIVATE cxx_std_11)
--- a/examples/beam-search/beam-search.cpp
+++ b/examples/beam-search/beam-search.cpp
@@ -1,188 +0,0 @@
-#include "common.h"
-#include "llama.h"
-
-#include <cassert>
-#include <cinttypes>
-#include <cmath>
-#include <cstdio>
-#include <cstring>
-#include <ctime>
-#include <fstream>
-#include <iostream>
-#include <string>
-#include <vector>
-
-#if defined (__unix__) || (defined (__APPLE__) && defined (__MACH__))
-#include <signal.h>
-#include <unistd.h>
-#elif defined (_WIN32)
-#define WIN32_LEAN_AND_MEAN
-#ifndef NOMINMAX
-#   define NOMINMAX
-#endif
-#include <windows.h>
-#include <signal.h>
-#endif
-
-// Used for debugging to print out beam tokens.
-struct ostream_beam_view {
-    llama_context * ctx;
-    llama_beam_view beam_view;
-};
-
-static std::ostream & operator<<(std::ostream & os, const ostream_beam_view & obv) {
-    os << "p(" << obv.beam_view.p << ") eob(" << std::boolalpha << obv.beam_view.eob << ") tokens(";
-    for (size_t i = 0 ; i < obv.beam_view.n_tokens ; ++i) {
-        os << llama_token_to_piece(obv.ctx, obv.beam_view.tokens[i]);
-    }
-    return os << ')';
-}
-
-// Put here anything you want back in beam_search_callback().
-struct beam_search_callback_data {
-    llama_context * ctx;
-    std::vector<llama_token> response;
-};
-
-// In this case, end-of-beam (eob) is equivalent to end-of-sentence (eos) but this need not always be the same.
-// For example, eob can be flagged due to maximum token length, stop words, etc.
-static bool is_at_eob(const beam_search_callback_data & callback_data, const llama_token * tokens, size_t n_tokens) {
-    return n_tokens && llama_token_is_eog(llama_get_model(callback_data.ctx), tokens[n_tokens-1]);
-}
-
-// Function matching type llama_beam_search_callback_fn_t.
-// Custom callback example is called each time the beams lengths increase:
-//  * Show progress by printing ',' following by number of convergent beam tokens if any.
-//  * When all beams converge to a common prefix, they are made available in beams_state.beams[0].
-//    This is also called when the stop condition is met.
-//    Collect tokens into std::vector<llama_token> response which is pointed to by callback_data.
-static void beam_search_callback(void * callback_data_ptr, llama_beams_state beams_state) {
-    auto& callback_data = *static_cast<beam_search_callback_data*>(callback_data_ptr);
-    // Mark beams as EOS as needed.
-    for (size_t i = 0 ; i < beams_state.n_beams ; ++i) {
-        llama_beam_view& beam_view = beams_state.beam_views[i];
-        if (!beam_view.eob && is_at_eob(callback_data, beam_view.tokens, beam_view.n_tokens)) {
-            beam_view.eob = true;
-        }
-    }
-    printf(",");  // Show progress
-    if (const size_t n = beams_state.common_prefix_length) {
-        callback_data.response.resize(callback_data.response.size() + n);
-        assert(0u < beams_state.n_beams);
-        const llama_token * tokens = beams_state.beam_views[0].tokens;
-        std::copy(tokens, tokens + n, callback_data.response.end() - n);
-        printf("%zu", n);
-    }
-    fflush(stdout);
-#if 1 // DEBUG: print current beams for this iteration
-    std::cout << "\n\nCurrent beams (last_call=" << beams_state.last_call << "):\n";
-    for (size_t i = 0 ; i < beams_state.n_beams ; ++i) {
-        std::cout << "beams["<<i<<"]: " << ostream_beam_view{callback_data.ctx,beams_state.beam_views[i]} << std::endl;
-    }
-#endif
-}
-
-int main(int argc, char ** argv)
-{
-    gpt_params params;
-    //params.n_gpu_layers = 200;
-
-    //---------------------------------
-    // Print help :
-    //---------------------------------
-
-    if ( argc < 2 || argv[1][0] == '-' )
-    {
-        printf( "Usage: %s MODEL_PATH [BEAM_WIDTH=2] [PROMPT]\n" , argv[0] );
-        return 1 ;
-    }
-
-    //---------------------------------
-    // Load parameters :
-    //---------------------------------
-
-    params.model = argv[1];
-
-    params.n_beams = 2 < argc ? std::stoi(argv[2]) : 2;
-
-    if ( argc > 3 )
-    {
-        params.prompt = argv[3];
-    }
-
-    if ( params.prompt.empty() )
-    {
-        params.prompt = "### Request:\nHow many countries are there?\n\n### Response:\n";
-    }
-
-    //---------------------------------
-    // Init LLM :
-    //---------------------------------
-
-    llama_backend_init();
-    llama_numa_init(params.numa);
-
-    llama_model * model;
-    llama_context * ctx;
-
-    std::tie(model, ctx) = llama_init_from_gpt_params( params );
-
-    if ( model == NULL )
-    {
-        fprintf( stderr , "%s: error: unable to load model\n" , __func__ );
-        return 1;
-    }
-
-    //---------------------------------
-    // Tokenize the prompt :
-    //---------------------------------
-
-    std::vector<llama_token> tokens_list = llama_tokenize(ctx, params.prompt, true);
-
-    const size_t max_context_size     = llama_n_ctx( ctx );
-    const size_t max_tokens_list_size = max_context_size - 4 ;
-
-    if (tokens_list.size() > max_tokens_list_size)
-    {
-        fprintf( stderr , "%s: error: prompt too long (%zu tokens, max %zu)\n" ,
-             __func__ , tokens_list.size() , max_tokens_list_size );
-        return 1;
-    }
-
-    fprintf( stderr, "\n\n" );
-
-    // Print the tokens from the prompt :
-
-    for( auto id : tokens_list )
-    {
-        std::cout << llama_token_to_piece(ctx, id);
-    }
-    std::cout << std::flush;
-
-    int n_past = 0;
-
-    if (llama_decode(ctx, llama_batch_get_one(tokens_list.data(), tokens_list.size(), n_past, 0)))
-    {
-        fprintf(stderr, "%s : failed to eval prompt.\n" , __func__ );
-        return 1;
-    }
-    n_past += tokens_list.size();
-
-    beam_search_callback_data callback_data{ctx, {}};
-    size_t const beam_width = static_cast<size_t>(params.n_beams);
-    int const n_predict = 256;
-    llama_beam_search(ctx, beam_search_callback, &callback_data, beam_width, n_past, n_predict);
-
-    std::cout << "\n\n";
-    for (llama_token const token_id : callback_data.response) {
-        std::cout << llama_token_to_piece(ctx,token_id);
-    }
-    std::cout << std::endl;
-
-    llama_free( ctx );
-    llama_free_model( model );
-
-    llama_backend_free();
-
-    return 0;
-}
--- a/examples/llama-bench/llama-bench.cpp
+++ b/examples/llama-bench/llama-bench.cpp
@@ -140,10 +140,11 @@ static std::string get_gpu_info() {
 }

 // command line params
-enum output_formats {CSV, JSON, MARKDOWN, SQL};
+enum output_formats {NONE, CSV, JSON, MARKDOWN, SQL};

 static const char * output_format_str(output_formats format) {
    switch (format) {
+        case NONE:     return "none";
        case CSV:      return "csv";
        case JSON:     return "json";
        case MARKDOWN: return "md";
@@ -152,6 +153,23 @@ static const char * output_format_str(output_formats format) {
    }
 }

+static bool output_format_from_str(const std::string & s, output_formats & format) {
+    if (s == "none") {
+        format = NONE;
+    } else if (s == "csv") {
+        format = CSV;
+    } else if (s == "json") {
+        format = JSON;
+    } else if (s == "md") {
+        format = MARKDOWN;
+    } else if (s == "sql") {
+        format = SQL;
+    } else {
+        return false;
+    }
+    return true;
+}
+
 static const char * split_mode_str(llama_split_mode mode) {
    switch (mode) {
        case LLAMA_SPLIT_MODE_NONE:  return "none";
@@ -190,31 +208,33 @@ struct cmd_params {
    int reps;
    bool verbose;
    output_formats output_format;
+    output_formats output_format_stderr;
 };

 static const cmd_params cmd_params_defaults = {
-    /* model         */ {"models/7B/ggml-model-q4_0.gguf"},
-    /* n_prompt      */ {512},
-    /* n_gen         */ {128},
-    /* n_pg          */ {},
-    /* n_batch       */ {2048},
-    /* n_ubatch      */ {512},
-    /* type_k        */ {GGML_TYPE_F16},
-    /* type_v        */ {GGML_TYPE_F16},
-    /* n_threads     */ {cpu_get_num_math()},
-    /* n_gpu_layers  */ {99},
-    /* rpc_servers   */ {""},
-    /* split_mode    */ {LLAMA_SPLIT_MODE_LAYER},
-    /* main_gpu      */ {0},
-    /* no_kv_offload */ {false},
-    /* flash_attn    */ {false},
-    /* tensor_split  */ {std::vector<float>(llama_max_devices(), 0.0f)},
-    /* use_mmap      */ {true},
-    /* embeddings    */ {false},
-    /* numa          */ GGML_NUMA_STRATEGY_DISABLED,
-    /* reps          */ 5,
-    /* verbose       */ false,
-    /* output_format */ MARKDOWN
+    /* model                */ {"models/7B/ggml-model-q4_0.gguf"},
+    /* n_prompt             */ {512},
+    /* n_gen                */ {128},
+    /* n_pg                 */ {},
+    /* n_batch              */ {2048},
+    /* n_ubatch             */ {512},
+    /* type_k               */ {GGML_TYPE_F16},
+    /* type_v               */ {GGML_TYPE_F16},
+    /* n_threads            */ {cpu_get_num_math()},
+    /* n_gpu_layers         */ {99},
+    /* rpc_servers          */ {""},
+    /* split_mode           */ {LLAMA_SPLIT_MODE_LAYER},
+    /* main_gpu             */ {0},
+    /* no_kv_offload        */ {false},
+    /* flash_attn           */ {false},
+    /* tensor_split         */ {std::vector<float>(llama_max_devices(), 0.0f)},
+    /* use_mmap             */ {true},
+    /* embeddings           */ {false},
+    /* numa                 */ GGML_NUMA_STRATEGY_DISABLED,
+    /* reps                 */ 5,
+    /* verbose              */ false,
+    /* output_format        */ MARKDOWN,
+    /* output_format_stderr */ NONE,
 };

 static void print_usage(int /* argc */, char ** argv) {
@@ -243,6 +263,7 @@ static void print_usage(int /* argc */, char ** argv) {
    printf("  -ts, --tensor-split <ts0/ts1/..>    (default: 0)\n");
    printf("  -r, --repetitions <n>               (default: %d)\n", cmd_params_defaults.reps);
    printf("  -o, --output <csv|json|md|sql>      (default: %s)\n", output_format_str(cmd_params_defaults.output_format));
+    printf("  -oe, --output-err <csv|json|md|sql> (default: %s)\n", output_format_str(cmd_params_defaults.output_format_stderr));
    printf("  -v, --verbose                       (default: %s)\n", cmd_params_defaults.verbose ? "1" : "0");
    printf("\n");
    printf("Multiple values can be given for each parameter by separating them with ',' or by specifying the parameter multiple times.\n");
@@ -284,6 +305,7 @@ static cmd_params parse_cmd_params(int argc, char ** argv) {

    params.verbose = cmd_params_defaults.verbose;
    params.output_format = cmd_params_defaults.output_format;
+    params.output_format_stderr = cmd_params_defaults.output_format_stderr;
    params.reps = cmd_params_defaults.reps;

    for (int i = 1; i < argc; i++) {
@@ -493,18 +515,13 @@ static cmd_params parse_cmd_params(int argc, char ** argv) {
                invalid_param = true;
                break;
            }
-            if (argv[i] == std::string("csv")) {
-                params.output_format = CSV;
-            } else if (argv[i] == std::string("json")) {
-                params.output_format = JSON;
-            } else if (argv[i] == std::string("md")) {
-                params.output_format = MARKDOWN;
-            } else if (argv[i] == std::string("sql")) {
-                params.output_format = SQL;
-            } else {
+            invalid_param = !output_format_from_str(argv[i], params.output_format);
+        } else if (arg == "-oe" || arg == "--output-err") {
+            if (++i >= argc) {
                invalid_param = true;
                break;
            }
+            invalid_param = !output_format_from_str(argv[i], params.output_format_stderr);
        } else if (arg == "-v" || arg == "--verbose") {
            params.verbose = true;
        } else {
@@ -1278,6 +1295,22 @@ static void llama_null_log_callback(enum ggml_log_level level, const char * text
    (void) user_data;
 }

+static std::unique_ptr<printer> create_printer(output_formats format) {
+    switch (format) {
+        case NONE:
+            return nullptr;
+        case CSV:
+            return std::unique_ptr<printer>(new csv_printer());
+        case JSON:
+            return std::unique_ptr<printer>(new json_printer());
+        case MARKDOWN:
+            return std::unique_ptr<printer>(new markdown_printer());
+        case SQL:
+            return std::unique_ptr<printer>(new sql_printer());
+    }
+    GGML_ASSERT(false);
+}
+
 int main(int argc, char ** argv) {
    // try to set locale for unicode characters in markdown
    setlocale(LC_CTYPE, ".UTF-8");
@@ -1304,26 +1337,18 @@ int main(int argc, char ** argv) {
    llama_numa_init(params.numa);

    // initialize printer
-    std::unique_ptr<printer> p;
-    switch (params.output_format) {
-        case CSV:
-            p.reset(new csv_printer());
-            break;
-        case JSON:
-            p.reset(new json_printer());
-            break;
-        case MARKDOWN:
-            p.reset(new markdown_printer());
-            break;
-        case SQL:
-            p.reset(new sql_printer());
-            break;
-        default:
-            assert(false);
-            exit(1);
+    std::unique_ptr<printer> p = create_printer(params.output_format);
+    std::unique_ptr<printer> p_err = create_printer(params.output_format_stderr);
+
+    if (p) {
+        p->fout = stdout;
+        p->print_header(params);
+    }
+
+    if (p_err) {
+        p_err->fout = stderr;
+        p_err->print_header(params);
    }
-    p->fout = stdout;
-    p->print_header(params);

    std::vector<cmd_params_instance> params_instances = get_cmd_params_instances(params);

@@ -1381,7 +1406,15 @@ int main(int argc, char ** argv) {
            t.samples_ns.push_back(t_ns);
        }

-        p->print_test(t);
+        if (p) {
+            p->print_test(t);
+            fflush(p->fout);
+        }
+
+        if (p_err) {
+            p_err->print_test(t);
+            fflush(p_err->fout);
+        }

        llama_print_timings(ctx);

@@ -1390,7 +1423,13 @@ int main(int argc, char ** argv) {

    llama_free_model(lmodel);

-    p->print_footer();
+    if (p) {
+        p->print_footer();
+    }
+
+    if (p_err) {
+        p_err->print_footer();
+    }

    llama_backend_free();

--- a/ggml-alloc.c
+++ b/ggml-alloc.c
@@ -750,7 +750,7 @@ static void ggml_gallocr_init_tensor(ggml_gallocr_t galloc, struct ggml_tensor *
                // this tensor was allocated without ggml-backend
                return;
            }
-            ggml_backend_view_init(galloc->buffers[buffer_id], tensor);
+            ggml_backend_view_init(tensor);
        }
    } else {
        if (tensor->data == NULL) {
@@ -899,12 +899,12 @@ static bool alloc_tensor_range(struct ggml_context * ctx,
            if (t->view_src == NULL) {
                ggml_tallocr_alloc(&tallocr, t);
            } else if (t->buffer == NULL) {
-                ggml_backend_view_init(buffer, t);
+                ggml_backend_view_init(t);
            }
        } else {
            if (t->view_src != NULL && t->buffer == NULL) {
                // view of a pre-allocated tensor
-                ggml_backend_view_init(buffer, t);
+                ggml_backend_view_init(t);
            }
        }
    }
--- a/ggml-backend.c
+++ b/ggml-backend.c
@@ -151,7 +151,7 @@ void ggml_backend_buffer_reset(ggml_backend_buffer_t buffer) {
 bool ggml_backend_buffer_copy_tensor(const struct ggml_tensor * src, struct ggml_tensor * dst) {
    ggml_backend_buffer_t dst_buf = dst->view_src ? dst->view_src->buffer : dst->buffer;
    if (dst_buf->iface.cpy_tensor) {
-        return src->buffer->iface.cpy_tensor(dst_buf, src, dst);
+        return dst_buf->iface.cpy_tensor(dst_buf, src, dst);
    }
    return false;
 }
@@ -1887,15 +1887,15 @@ ggml_backend_t ggml_backend_sched_get_tensor_backend(ggml_backend_sched_t sched,

 // utils

-void ggml_backend_view_init(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor) {
+void ggml_backend_view_init(struct ggml_tensor * tensor) {
    GGML_ASSERT(tensor->buffer == NULL);
    GGML_ASSERT(tensor->view_src != NULL);
    GGML_ASSERT(tensor->view_src->buffer != NULL);
    GGML_ASSERT(tensor->view_src->data != NULL);

-    tensor->buffer = buffer;
+    tensor->buffer = tensor->view_src->buffer;
    tensor->data = (char *)tensor->view_src->data + tensor->view_offs;
-    ggml_backend_buffer_init_tensor(buffer, tensor);
+    ggml_backend_buffer_init_tensor(tensor->buffer, tensor);
 }

 void ggml_backend_tensor_alloc(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor, void * addr) {
@@ -1954,7 +1954,7 @@ static void graph_copy_init_tensor(struct ggml_hash_set hash_set, struct ggml_te
    struct ggml_tensor * dst = node_copies[id];
    if (dst->view_src != NULL) {
        graph_copy_init_tensor(hash_set, node_copies, node_init, src->view_src);
-        ggml_backend_view_init(dst->view_src->buffer, dst);
+        ggml_backend_view_init(dst);
    }
    else {
        ggml_backend_tensor_copy(src, dst);
--- a/ggml-backend.h
+++ b/ggml-backend.h
@@ -225,7 +225,7 @@ extern "C" {

    // Tensor initialization
    GGML_API void ggml_backend_tensor_alloc(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor, void * addr);
-    GGML_API void ggml_backend_view_init(ggml_backend_buffer_t buffer, struct ggml_tensor * tensor);
+    GGML_API void ggml_backend_view_init(struct ggml_tensor * tensor);


 #ifdef  __cplusplus
--- a/ggml-rpc.cpp
+++ b/ggml-rpc.cpp
@@ -491,7 +491,7 @@ GGML_CALL static ggml_backend_buffer_t ggml_backend_rpc_buffer_type_alloc_buffer
    if (remote_ptr != 0) {
        ggml_backend_buffer_t buffer = ggml_backend_buffer_init(buft,
            ggml_backend_rpc_buffer_interface,
-            new ggml_backend_rpc_buffer_context{sock, {}, remote_ptr, "RPC"},
+            new ggml_backend_rpc_buffer_context{sock, {}, remote_ptr, "RPC[" + std::string(buft_ctx->endpoint) + "]"},
            remote_size);
        return buffer;
    } else {
@@ -692,7 +692,7 @@ GGML_API GGML_CALL ggml_backend_buffer_type_t ggml_backend_rpc_buffer_type(const
 GGML_CALL ggml_backend_t ggml_backend_rpc_init(const char * endpoint) {
    ggml_backend_rpc_context * ctx = new ggml_backend_rpc_context {
        /* .endpoint  = */ endpoint,
-        /* .name      = */ "RPC",
+        /* .name      = */ "RPC[" + std::string(endpoint) + "]",
    };

    ggml_backend_t backend = new ggml_backend {
--- a/ggml-vulkan-shaders.hpp
+++ b/ggml-vulkan-shaders.hpp
--- a/ggml-vulkan.cpp
+++ b/ggml-vulkan.cpp
--- a/ggml.c
+++ b/ggml.c
@@ -5,6 +5,7 @@
 #include "ggml-quants.h"
 #include "ggml.h"

+
 #if defined(_MSC_VER) || defined(__MINGW32__)
 #include <malloc.h> // using malloc.h with MSC/MINGW
 #elif !defined(__FreeBSD__) && !defined(__NetBSD__) && !defined(__OpenBSD__)
@@ -28,6 +29,10 @@
 #include <syscall.h>
 #endif

+#ifdef GGML_USE_OPENMP
+#include <omp.h>
+#endif
+
 #ifdef GGML_USE_METAL
 #include <unistd.h>
 #endif
@@ -1756,7 +1761,7 @@ struct ggml_compute_state_shared {
    int64_t perf_node_start_cycles;
    int64_t perf_node_start_time_us;

-    const int n_threads;
+    int n_threads;

    // synchronization primitives
    atomic_int n_active;  // num active threads
@@ -2267,6 +2272,11 @@ inline static float ggml_silu_f32(float x) {
    return x/(1.0f + expf(-x));
 }

+#if __FINITE_MATH_ONLY__
+#error "some routines in ggml.c require non-finite math arithmetics -- pass -fno-finite-math-only to the compiler to fix"
+#error "ref: https://github.com/ggerganov/llama.cpp/pull/7154#issuecomment-2143844461"
+#endif
+
 #if defined(__ARM_NEON) && defined(__aarch64__)

 // adapted from arm limited optimized routine
@@ -19670,6 +19680,59 @@ struct ggml_cplan ggml_graph_plan(const struct ggml_cgraph * cgraph, int n_threa
    return cplan;
 }

+static enum ggml_status ggml_graph_compute_parallel(struct ggml_compute_state * workers, int n_threads) {
+    enum ggml_status compute_status = GGML_STATUS_SUCCESS;
+
+#ifdef GGML_USE_OPENMP
+    if (n_threads > 1) {
+        #pragma omp parallel num_threads(n_threads)
+        {
+            #pragma omp single
+            {
+                // update the number of threads from the actual number of threads that we got from OpenMP
+                n_threads = omp_get_num_threads();
+                workers[0].shared->n_threads = n_threads;
+                workers[0].shared->n_active  = n_threads;
+            }
+            ggml_graph_compute_thread(&workers[omp_get_thread_num()]);
+        }
+    } else {
+        ggml_graph_compute_thread(&workers[0]);
+    }
+#else
+    // create thread pool
+    if (n_threads > 1) {
+        for (int j = 1; j < n_threads; ++j) {
+            const int rc = ggml_thread_create(&workers[j].thrd, NULL, ggml_graph_compute_thread, &workers[j]);
+            GGML_ASSERT(rc == 0);
+            UNUSED(rc);
+        }
+    }
+
+    // this is a work thread too
+    ggml_graph_compute_thread(&workers[0]);
+
+    // join or kill thread pool
+    if (n_threads > 1) {
+        for (int j = 1; j < n_threads; j++) {
+            const int rc = ggml_thread_join(workers[j].thrd, NULL);
+            GGML_ASSERT(rc == 0);
+            UNUSED(rc);
+        }
+    }
+#endif
+    // don't leave affinity set on the main thread
+    clear_numa_thread_affinity();
+
+    for (int j = 0; j < n_threads; j++) {
+        if (workers[j].ec != GGML_STATUS_SUCCESS) {
+            compute_status = workers[j].ec;
+            break;
+        }
+    }
+    return compute_status;
+}
+
 enum ggml_status ggml_graph_compute(struct ggml_cgraph * cgraph, struct ggml_cplan * cplan) {
    {
        GGML_ASSERT(cplan);
@@ -19680,7 +19743,11 @@ enum ggml_status ggml_graph_compute(struct ggml_cgraph * cgraph, struct ggml_cpl
        }
    }

-    const int n_threads = cplan->n_threads;
+    int n_threads = cplan->n_threads;
+
+#if defined(GGML_USE_OPENMP)
+    n_threads = MIN(n_threads, omp_get_max_threads());
+#endif

    struct ggml_compute_state_shared state_shared = {
        /*.cgraph                  =*/ cgraph,
@@ -19696,47 +19763,20 @@ enum ggml_status ggml_graph_compute(struct ggml_cgraph * cgraph, struct ggml_cpl
        /*.current_chunk;          =*/ 0,
    };
    struct ggml_compute_state * workers = alloca(sizeof(struct ggml_compute_state)*n_threads);
-
-    // create thread pool
-    if (n_threads > 1) {
-        for (int j = 1; j < n_threads; ++j) {
-            workers[j] = (struct ggml_compute_state) {
-                .thrd   = 0,
-                .ith = j,
-                .shared = &state_shared,
-                .ec = GGML_STATUS_SUCCESS,
-            };
-
-            const int rc = ggml_thread_create(&workers[j].thrd, NULL, ggml_graph_compute_thread, &workers[j]);
-            GGML_ASSERT(rc == 0);
-            UNUSED(rc);
-        }
-    }
-
-    workers[0].ith = 0;
-    workers[0].shared = &state_shared;
-    workers[0].ec = GGML_STATUS_SUCCESS;
-
    const int64_t perf_start_cycles  = ggml_perf_cycles();
    const int64_t perf_start_time_us = ggml_perf_time_us();

-    // this is a work thread too
-    ggml_graph_compute_thread(&workers[0]);
-    enum ggml_status compute_status = workers[0].ec;
-
-    // don't leave affinity set on the main thread
-    clear_numa_thread_affinity();
-
-    // join or kill thread pool
-    if (n_threads > 1) {
-        for (int j = 1; j < n_threads; j++) {
-            const int rc = ggml_thread_join(workers[j].thrd, NULL);
-            GGML_ASSERT(rc == 0);
-            if (workers[j].ec != GGML_STATUS_SUCCESS)
-                compute_status = workers[j].ec;
-        }
+    for (int j = 0; j < n_threads; ++j) {
+        workers[j] = (struct ggml_compute_state) {
+            .thrd   = 0,
+            .ith    = j,
+            .shared = &state_shared,
+            .ec     = GGML_STATUS_SUCCESS,
+        };
    }

+    enum ggml_status compute_status = ggml_graph_compute_parallel(workers, n_threads);
+
    // performance stats (graph)
    {
        int64_t perf_cycles_cur  = ggml_perf_cycles()  - perf_start_cycles;
--- a/ggml_vk_generate_shaders.py
+++ b/ggml_vk_generate_shaders.py
@@ -225,10 +225,7 @@ mulmat_head = """#version 450
 #extension GL_EXT_shader_16bit_storage : require

 #ifdef MUL_MAT_ID
-#extension GL_EXT_buffer_reference2 : require
-#extension GL_EXT_nonuniform_qualifier : require
-#extension GL_EXT_scalar_block_layout : require
-#extension GL_EXT_shader_explicit_arithmetic_types_int8 : require
+#extension GL_EXT_shader_explicit_arithmetic_types_int16 : require

 #define EXPERT_COUNT 8
 #endif
@@ -260,30 +257,22 @@ layout (push_constant) uniform parameter
    uint stride_a;
    uint stride_b;
    uint stride_d;
-    uint k_split;
-
-    uint ne02;
-    uint ne12;
-    uint broadcast2;
-    uint broadcast3;

    uint batch_stride_a;
    uint batch_stride_b;
    uint batch_stride_d;

 #ifdef MUL_MAT_ID
-    uint expert_stride_a;
-    uint expert_stride_b0;
-    uint expert_stride_b1;
-    uint expert_stride_d;
-
-    uint ids_stride;
-
-    uint n_as;
    uint nei0;
    uint nei1;
    uint nbi1;
    uint ne11;
+#else
+    uint k_split;
+    uint ne02;
+    uint ne12;
+    uint broadcast2;
+    uint broadcast3;
 #endif
 } p;

@@ -301,16 +290,14 @@ shared FLOAT_TYPE buf_a[BM * (BK+1)];
 shared FLOAT_TYPE buf_b[BN * (BK+1)];

 #ifdef MUL_MAT_ID
-shared u8vec2 rowids[2048];
+shared u16vec2 row_ids[2048];
 #endif

 void main() {
 #ifdef MUL_MAT_ID
-    const uint batch_idx = gl_GlobalInvocationID.z / p.n_as;
-    const uint expert_idx = gl_GlobalInvocationID.z % p.n_as;
+    const uint expert_idx = gl_GlobalInvocationID.z;
 #else
    const uint batch_idx = gl_GlobalInvocationID.z;
-#endif

    const uint i13 = batch_idx / p.ne12;
    const uint i12 = batch_idx % p.ne12;
@@ -319,6 +306,7 @@ void main() {
    const uint i02 = i12 / p.broadcast2;

    const uint batch_idx_a = i03 * p.ne02 + i02;
+#endif

    const uint blocks_m = (p.M + BM - 1) / BM;
    const uint ir = gl_WorkGroupID.x % blocks_m;
@@ -350,30 +338,38 @@ void main() {
    for (uint ii1 = 0; ii1 < p.nei1; ii1++) {
        for (uint ii0 = 0; ii0 < p.nei0; ii0++) {
            if (data_ids[ii1*p.nbi1 + ii0] == expert_idx) {
-                rowids[_ne1] = u8vec2(ii0, ii1);
+                row_ids[_ne1] = u16vec2(ii0, ii1);
                _ne1++;
            }
        }
    }

-    const u8vec2 id = rowids[ir * BN + ic];
+    barrier();
+
+    // Workgroup has no work
+    if (ic * BN >= _ne1) return;
 #endif

+#ifdef MUL_MAT_ID
+    const uint start_k = 0;
+    const uint end_k = p.K;
+#else
    const uint start_k = ik * p.k_split;
    const uint end_k = min(p.K, (ik + 1) * p.k_split);
+#endif

    uint pos_a = (
 #ifdef MUL_MAT_ID
-        expert_idx * p.expert_stride_a +
+        expert_idx * p.batch_stride_a +
+#else
+        batch_idx_a * p.batch_stride_a +
 #endif
-        batch_idx_a * p.batch_stride_a + ir * BM * p.stride_a + start_k) / LOAD_VEC_A;
-    uint pos_b = (
+        ir * BM * p.stride_a + start_k) / LOAD_VEC_A;
 #ifdef MUL_MAT_ID
-        id.y * p.expert_stride_b1 +
-        (id.x % p.ne11) * p.expert_stride_b0 +
+    uint pos_b = 0;
+#else
+    uint pos_b = (batch_idx * p.batch_stride_b + ic * BN * p.stride_b + start_k) / LOAD_VEC_B;
 #endif
-        batch_idx * p.batch_stride_b +
-        ic * BN * p.stride_b + start_k) / LOAD_VEC_B;

    float sums[WMITER * TM * WNITER * TN];
    FLOAT_TYPE cache_a[WMITER * TM];
@@ -620,7 +616,12 @@ mulmat_body2 = """
        }
        [[unroll]] for (uint l = 0; l < BN; l += loadstride_b) {
 #if LOAD_VEC_B == 8
+#ifdef MUL_MAT_ID
+            const u16vec2 row_idx = row_ids[ic * BN + loadc_b + l];
+            const uint idx = pos_b + row_idx.y * p.batch_stride_b / LOAD_VEC_B + (row_idx.x % p.ne11) * p.stride_b / LOAD_VEC_B + loadr_b;
+#else
            const uint idx = pos_b + (loadc_b + l) * p.stride_b / LOAD_VEC_B + loadr_b;
+#endif
            const uint buf_idx = (loadc_b + l) * (BK+1) + loadr_b * LOAD_VEC_B;
            buf_b[buf_idx + 0] = FLOAT_TYPE(data_b[idx][0].x);
            buf_b[buf_idx + 1] = FLOAT_TYPE(data_b[idx][0].y);
@@ -631,18 +632,31 @@ mulmat_body2 = """
            buf_b[buf_idx + 6] = FLOAT_TYPE(data_b[idx][1].z);
            buf_b[buf_idx + 7] = FLOAT_TYPE(data_b[idx][1].w);
 #elif LOAD_VEC_B == 4
+#ifdef MUL_MAT_ID
+            const u16vec2 row_idx = row_ids[ic * BN + loadc_b + l];
+            const uint idx = pos_b + row_idx.y * p.batch_stride_b / LOAD_VEC_B + (row_idx.x % p.ne11) * p.stride_b / LOAD_VEC_B + loadr_b;
+#else
            const uint idx = pos_b + (loadc_b + l) * p.stride_b / LOAD_VEC_B + loadr_b;
+#endif
            const uint buf_idx = (loadc_b + l) * (BK+1) + loadr_b * LOAD_VEC_B;
            buf_b[buf_idx + 0] = FLOAT_TYPE(data_b[idx].x);
            buf_b[buf_idx + 1] = FLOAT_TYPE(data_b[idx].y);
            buf_b[buf_idx + 2] = FLOAT_TYPE(data_b[idx].z);
            buf_b[buf_idx + 3] = FLOAT_TYPE(data_b[idx].w);
-#else
+#elif !MUL_MAT_ID
            if (ic * BN + loadc_b + l < p.N && block + loadr_b < end_k) {
                buf_b[(loadc_b + l) * (BK+1) + loadr_b] = FLOAT_TYPE(data_b[pos_b + (loadc_b + l) * p.stride_b + loadr_b]);
            } else {
                buf_b[(loadc_b + l) * (BK+1) + loadr_b] = FLOAT_TYPE(0.0f);
            }
+#else
+            const uint row_i = ic * BN + loadc_b + l;
+            if (row_i < _ne1) {
+                const u16vec2 row_idx = row_ids[row_i];
+                buf_b[(loadc_b + l) * (BK+1) + loadr_b] = FLOAT_TYPE(data_b[pos_b + row_idx.y * p.batch_stride_b + (row_idx.x % p.ne11) * p.stride_b + loadr_b]);
+            } else {
+                buf_b[(loadc_b + l) * (BK+1) + loadr_b] = FLOAT_TYPE(0.0f);
+            }
 #endif
        }

@@ -681,11 +695,9 @@ mulmat_body2 = """
    const uint dr = ir * BM + warp_r * WM;
    const uint dc = ic * BN + warp_c * WN;

-    const uint offsets =
-#ifdef MUL_MAT_ID
-            expert_idx * p.expert_stride_d +
+#ifndef MUL_MAT_ID
+    const uint offsets = batch_idx * p.batch_stride_d + ik * p.batch_stride_d * gl_NumWorkGroups.z;
 #endif
-            batch_idx * p.batch_stride_d + ik * p.batch_stride_d * gl_NumWorkGroups.z;

    [[unroll]] for (uint wsic = 0; wsic < WNITER; wsic++) {
        [[unroll]] for (uint wsir = 0; wsir < WMITER; wsir++) {
@@ -693,10 +705,20 @@ mulmat_body2 = """
            const uint dr_warp = dr + wsir * WSUBM + tiwr * TM;
            const uint dc_warp = dc + wsic * WSUBN + tiwc * TN;
            [[unroll]] for (uint cc = 0; cc < TN; cc++) {
+#ifdef MUL_MAT_ID
+                const uint row_i = dc_warp + cc;
+                if (row_i >= _ne1) break;
+
+                const u16vec2 row_idx = row_ids[row_i];
+#endif
                [[unroll]] for (uint cr = 0; cr < TM; cr++) {
+#ifdef MUL_MAT_ID
+                    data_d[row_idx.y * p.batch_stride_d + row_idx.x * p.stride_d + dr_warp + cr] = D_TYPE(sums[(wsic * TN + cc) * (WMITER * TM) + wsir * TM + cr]);
+#else
                    if (dr_warp + cr < p.M && dc_warp + cc < p.N) {
                        data_d[offsets + (dc_warp + cc) * p.stride_d + dr_warp + cr] = D_TYPE(sums[(wsic * TN + cc) * (WMITER * TM) + wsir * TM + cr]);
                    }
+#endif
                }
            }
        }
@@ -1172,28 +1194,59 @@ layout (push_constant) uniform parameter
    uint stride_b;
    uint stride_d;

-    uint ne02;
-    uint ne12;
-    uint broadcast2;
-    uint broadcast3;
-
    uint batch_stride_a;
    uint batch_stride_b;
    uint batch_stride_d;

 #ifdef MUL_MAT_ID
-    uint expert_stride_a;
-    uint expert_stride_b0;
-    uint expert_stride_b1;
-    uint expert_stride_d0;
-    uint expert_stride_d1;
-
-    uint ne11;
    uint nei0;
-    uint nbi1;
-    uint n_as;
+    uint ne11;
+#else
+    uint ne02;
+    uint ne12;
+    uint broadcast2;
+    uint broadcast3;
 #endif
 } p;
+
+void get_offsets(out uint a_offset, out uint b_offset, out uint d_offset) {
+#ifdef MUL_MAT_ID
+    const uint expert_idx = gl_GlobalInvocationID.y;
+#else
+    const uint batch_idx = gl_GlobalInvocationID.y;
+#endif
+
+#ifndef MUL_MAT_ID
+    const uint i13 = batch_idx / p.ne12;
+    const uint i12 = batch_idx % p.ne12;
+
+    const uint i03 = i13 / p.broadcast3;
+    const uint i02 = i12 / p.broadcast2;
+
+    const uint batch_idx_a = i03 * p.ne02 + i02;
+#else
+    const uint expert_id = data_ids[expert_idx];
+#endif
+
+    a_offset =
+#ifdef MUL_MAT_ID
+            expert_id * p.batch_stride_a;
+#else
+            batch_idx_a * p.batch_stride_a;
+#endif
+    b_offset =
+#ifdef MUL_MAT_ID
+            (expert_idx % p.ne11) * p.stride_b;
+#else
+            batch_idx * p.batch_stride_b;
+#endif
+    d_offset =
+#ifdef MUL_MAT_ID
+            expert_idx * p.stride_d;
+#else
+            batch_idx * p.batch_stride_d;
+#endif
+}
 """

 mul_mat_vec_body = """
@@ -1206,41 +1259,9 @@ shared FLOAT_TYPE tmp[BLOCK_SIZE];
 void main() {
    const uint row = gl_WorkGroupID.x;
    const uint tid = gl_LocalInvocationID.x;
-    const uint batch_idx = gl_GlobalInvocationID.y;
-#ifdef MUL_MAT_ID
-    const uint expert_idx1 = gl_GlobalInvocationID.z / p.nei0;
-    const uint expert_idx0 = gl_GlobalInvocationID.z % p.nei0;
-#endif

-    const uint i13 = batch_idx / p.ne12;
-    const uint i12 = batch_idx % p.ne12;
-
-    const uint i03 = i13 / p.broadcast3;
-    const uint i02 = i12 / p.broadcast2;
-
-    const uint batch_idx_a = i03 * p.ne02 + i02;
-
-#ifdef MUL_MAT_ID
-    const uint expert_id = data_ids[expert_idx1 * p.nbi1 + expert_idx0];
-#endif
-
-    const uint a_offset =
-#ifdef MUL_MAT_ID
-            expert_id * p.expert_stride_a +
-#endif
-            batch_idx_a * p.batch_stride_a;
-    const uint b_offset =
-#ifdef MUL_MAT_ID
-            (expert_idx0 % p.ne11) * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_b;
-    const uint d_offset =
-#ifdef MUL_MAT_ID
-            expert_idx0 * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_d;
+    uint a_offset, b_offset, d_offset;
+    get_offsets(a_offset, b_offset, d_offset);

    const uint y_offset = QUANT_R == 1 ? 1 : QUANT_K/2;

@@ -1281,41 +1302,9 @@ shared FLOAT_TYPE tmp[32];

 void main() {
    const uint row = gl_WorkGroupID.x;
-    const uint batch_idx = gl_GlobalInvocationID.y;
-#ifdef MUL_MAT_ID
-    const uint expert_idx1 = gl_GlobalInvocationID.z / p.nei0;
-    const uint expert_idx0 = gl_GlobalInvocationID.z % p.nei0;
-#endif

-    const uint i13 = batch_idx / p.ne12;
-    const uint i12 = batch_idx % p.ne12;
-
-    const uint i03 = i13 / p.broadcast3;
-    const uint i02 = i12 / p.broadcast2;
-
-    const uint batch_idx_a = i03 * p.ne02 + i02;
-
-#ifdef MUL_MAT_ID
-    const uint expert_id = data_ids[expert_idx1 * p.nbi1 + expert_idx0];
-#endif
-
-    const uint a_offset =
-#ifdef MUL_MAT_ID
-            expert_id * p.expert_stride_a +
-#endif
-            batch_idx_a * p.batch_stride_a;
-    const uint b_offset =
-#ifdef MUL_MAT_ID
-            (expert_idx0 % p.ne11) * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_b;
-    const uint d_offset =
-#ifdef MUL_MAT_ID
-            expert_idx0 * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_d;
+    uint a_offset, b_offset, d_offset;
+    get_offsets(a_offset, b_offset, d_offset);

    const uint num_blocks_per_row = p.ncols / QUANT_K;
    const uint ib0 = a_offset / QUANT_K + row*num_blocks_per_row;
@@ -1384,41 +1373,9 @@ shared FLOAT_TYPE tmp[32];

 void main() {
    const uint row = gl_WorkGroupID.x;
-    const uint batch_idx = gl_GlobalInvocationID.y;
-#ifdef MUL_MAT_ID
-    const uint expert_idx1 = gl_GlobalInvocationID.z / p.nei0;
-    const uint expert_idx0 = gl_GlobalInvocationID.z % p.nei0;
-#endif

-    const uint i13 = batch_idx / p.ne12;
-    const uint i12 = batch_idx % p.ne12;
-
-    const uint i03 = i13 / p.broadcast3;
-    const uint i02 = i12 / p.broadcast2;
-
-    const uint batch_idx_a = i03 * p.ne02 + i02;
-
-#ifdef MUL_MAT_ID
-    const uint expert_id = data_ids[expert_idx1 * p.nbi1 + expert_idx0];
-#endif
-
-    const uint a_offset =
-#ifdef MUL_MAT_ID
-            expert_id * p.expert_stride_a +
-#endif
-            batch_idx_a * p.batch_stride_a;
-    const uint b_offset =
-#ifdef MUL_MAT_ID
-            (expert_idx0 % p.ne11) * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_b;
-    const uint d_offset =
-#ifdef MUL_MAT_ID
-            expert_idx0 * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_d;
+    uint a_offset, b_offset, d_offset;
+    get_offsets(a_offset, b_offset, d_offset);

    const uint num_blocks_per_row = p.ncols / QUANT_K;
    const uint ib0 = a_offset / QUANT_K + row*num_blocks_per_row;
@@ -1480,41 +1437,9 @@ shared FLOAT_TYPE tmp[32];

 void main() {
    const uint row = gl_WorkGroupID.x;
-    const uint batch_idx = gl_GlobalInvocationID.y;
-#ifdef MUL_MAT_ID
-    const uint expert_idx1 = gl_GlobalInvocationID.z / p.nei0;
-    const uint expert_idx0 = gl_GlobalInvocationID.z % p.nei0;
-#endif

-    const uint i13 = batch_idx / p.ne12;
-    const uint i12 = batch_idx % p.ne12;
-
-    const uint i03 = i13 / p.broadcast3;
-    const uint i02 = i12 / p.broadcast2;
-
-    const uint batch_idx_a = i03 * p.ne02 + i02;
-
-#ifdef MUL_MAT_ID
-    const uint expert_id = data_ids[expert_idx1 * p.nbi1 + expert_idx0];
-#endif
-
-    const uint a_offset =
-#ifdef MUL_MAT_ID
-            expert_id * p.expert_stride_a +
-#endif
-            batch_idx_a * p.batch_stride_a;
-    const uint b_offset =
-#ifdef MUL_MAT_ID
-            (expert_idx0 % p.ne11) * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_b;
-    const uint d_offset =
-#ifdef MUL_MAT_ID
-            expert_idx0 * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_d;
+    uint a_offset, b_offset, d_offset;
+    get_offsets(a_offset, b_offset, d_offset);

    const uint num_blocks_per_row = p.ncols / QUANT_K;
    const uint ib0 = a_offset / QUANT_K + row*num_blocks_per_row;
@@ -1625,41 +1550,9 @@ shared FLOAT_TYPE tmp[32];

 void main() {
    const uint row = gl_WorkGroupID.x;
-    const uint batch_idx = gl_GlobalInvocationID.y;
-#ifdef MUL_MAT_ID
-    const uint expert_idx1 = gl_GlobalInvocationID.z / p.nei0;
-    const uint expert_idx0 = gl_GlobalInvocationID.z % p.nei0;
-#endif

-    const uint i13 = batch_idx / p.ne12;
-    const uint i12 = batch_idx % p.ne12;
-
-    const uint i03 = i13 / p.broadcast3;
-    const uint i02 = i12 / p.broadcast2;
-
-    const uint batch_idx_a = i03 * p.ne02 + i02;
-
-#ifdef MUL_MAT_ID
-    const uint expert_id = data_ids[expert_idx1 * p.nbi1 + expert_idx0];
-#endif
-
-    const uint a_offset =
-#ifdef MUL_MAT_ID
-            expert_id * p.expert_stride_a +
-#endif
-            batch_idx_a * p.batch_stride_a;
-    const uint b_offset =
-#ifdef MUL_MAT_ID
-            (expert_idx0 % p.ne11) * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_b;
-    const uint d_offset =
-#ifdef MUL_MAT_ID
-            expert_idx0 * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_d;
+    uint a_offset, b_offset, d_offset;
+    get_offsets(a_offset, b_offset, d_offset);

    const uint num_blocks_per_row = p.ncols / QUANT_K;
    const uint ib0 = a_offset / QUANT_K + row*num_blocks_per_row;
@@ -1766,41 +1659,9 @@ shared FLOAT_TYPE tmp[32];

 void main() {
    const uint row = gl_WorkGroupID.x;
-    const uint batch_idx = gl_GlobalInvocationID.y;
-#ifdef MUL_MAT_ID
-    const uint expert_idx1 = gl_GlobalInvocationID.z / p.nei0;
-    const uint expert_idx0 = gl_GlobalInvocationID.z % p.nei0;
-#endif

-    const uint i13 = batch_idx / p.ne12;
-    const uint i12 = batch_idx % p.ne12;
-
-    const uint i03 = i13 / p.broadcast3;
-    const uint i02 = i12 / p.broadcast2;
-
-    const uint batch_idx_a = i03 * p.ne02 + i02;
-
-#ifdef MUL_MAT_ID
-    const uint expert_id = data_ids[expert_idx1 * p.nbi1 + expert_idx0];
-#endif
-
-    const uint a_offset =
-#ifdef MUL_MAT_ID
-            expert_id * p.expert_stride_a +
-#endif
-            batch_idx_a * p.batch_stride_a;
-    const uint b_offset =
-#ifdef MUL_MAT_ID
-            (expert_idx0 % p.ne11) * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_b;
-    const uint d_offset =
-#ifdef MUL_MAT_ID
-            expert_idx0 * p.expert_stride_b0 +
-            expert_idx1 * p.expert_stride_b1 +
-#endif
-            batch_idx * p.batch_stride_d;
+    uint a_offset, b_offset, d_offset;
+    get_offsets(a_offset, b_offset, d_offset);

    const uint num_blocks_per_row = p.ncols / QUANT_K;
    const uint ib0 = a_offset / QUANT_K + row*num_blocks_per_row;
@@ -2143,12 +2004,18 @@ void main() {

 generic_binary_op_combined = f"{generic_binary_op_head}\n{generic_binary_op_layout}\n{generic_binary_op_funcs}\n{generic_binary_op_main}"

-# MUL F32
+# MUL
 mul_body = """
    data_d[p.d_offset + dst_idx(gl_GlobalInvocationID.x)] = D_TYPE(FLOAT_TYPE(data_a[src0_idx(gl_GlobalInvocationID.x)]) * FLOAT_TYPE(data_b[src1_idx(gl_GlobalInvocationID.x)]));
 }
 """

+# DIV
+div_body = """
+    data_d[p.d_offset + dst_idx(gl_GlobalInvocationID.x)] = D_TYPE(FLOAT_TYPE(data_a[src0_idx(gl_GlobalInvocationID.x)]) / FLOAT_TYPE(data_b[src1_idx(gl_GlobalInvocationID.x)]));
+}
+"""
+
 # ADD
 add_body = """
    data_d[p.d_offset + dst_idx(gl_GlobalInvocationID.x)] = D_TYPE(FLOAT_TYPE(data_a[src0_idx(gl_GlobalInvocationID.x)]) + FLOAT_TYPE(data_b[src1_idx(gl_GlobalInvocationID.x)]));
@@ -2759,6 +2626,41 @@ void main() {
 }
 """

+sum_rows_src = """
+#extension GL_EXT_control_flow_attributes : enable
+layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;
+
+layout (binding = 0) readonly buffer A {A_TYPE data_a[];};
+layout (binding = 1) writeonly buffer D {D_TYPE data_d[];};
+
+layout (constant_id = 0) const uint BLOCK_SIZE = 32;
+
+shared FLOAT_TYPE tmp[BLOCK_SIZE];
+
+void main() {
+    const uint row = gl_WorkGroupID.x;
+    const uint col = gl_LocalInvocationID.x;
+
+    tmp[col] = FLOAT_TYPE(0.0f);
+
+    for (uint i = col; i < p.KX; i += BLOCK_SIZE) {
+        tmp[col] += FLOAT_TYPE(data_a[row*p.KX + i]);
+    }
+
+    barrier();
+    [[unroll]] for (int s = int(BLOCK_SIZE) / 2; s > 0; s >>= 1) {
+        if (col < s) {
+            tmp[col] += tmp[col + s];
+        }
+        barrier();
+    }
+
+    if (col == 0) {
+        data_d[row] = D_TYPE(tmp[0]);
+    }
+}
+"""
+
 GLSLC = "glslc"

 VK_NUM_TYPES = 16
@@ -2940,66 +2842,66 @@ async def main():
        tasks.append(string_to_spv("matmul_q6_k_f32_aligned", "".join(stream), {"LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q6_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

        # MUL_MAT_ID
-        # stream.clear()
-        # stream.extend((mulmat_head, shader_float_type, mulmat_body1, mulmat_load_scalar, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_f32", "".join(stream), {"MUL_MAT_ID": "1", "A_TYPE": "float", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": load_vec, "LOAD_VEC_B": load_vec, "A_TYPE": vec_type, "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_float_type, mulmat_body1, mulmat_load_scalar, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_f32", "".join(stream), {"MUL_MAT_ID": "1", "A_TYPE": "float", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": load_vec, "LOAD_VEC_B": load_vec, "A_TYPE": vec_type, "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # tasks.append(string_to_spv("matmul_id_f16", "".join(stream), {"MUL_MAT_ID": "1", "A_TYPE": "float16_t", "B_TYPE": "float16_t", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_f16_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": load_vec, "LOAD_VEC_B": load_vec, "A_TYPE": vec_type_f16, "B_TYPE": vec_type_f16, "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_f16", "".join(stream), {"MUL_MAT_ID": "1", "A_TYPE": "float16_t", "B_TYPE": "float16_t", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_f16_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": load_vec, "LOAD_VEC_B": load_vec, "A_TYPE": vec_type_f16, "B_TYPE": vec_type_f16, "D_TYPE": "float"}, fp16))

-        # tasks.append(string_to_spv("matmul_id_f16_f32", "".join(stream), {"MUL_MAT_ID": "1", "A_TYPE": "float16_t", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_f16_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": load_vec, "LOAD_VEC_B": load_vec, "A_TYPE": vec_type_f16, "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_f16_f32", "".join(stream), {"MUL_MAT_ID": "1", "A_TYPE": "float16_t", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_f16_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": load_vec, "LOAD_VEC_B": load_vec, "A_TYPE": vec_type_f16, "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # stream.clear()
-        # stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q4_0_defines, mulmat_body1, mulmat_load_q4_0, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_q4_0_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q4_0", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_q4_0_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q4_0", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q4_0_defines, mulmat_body1, mulmat_load_q4_0, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_q4_0_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q4_0", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_q4_0_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q4_0", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # stream.clear()
-        # stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q4_1_defines, mulmat_body1, mulmat_load_q4_1, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_q4_1_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q4_1", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_q4_1_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q4_1", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q4_1_defines, mulmat_body1, mulmat_load_q4_1, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_q4_1_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q4_1", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_q4_1_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q4_1", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # stream.clear()
-        # stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q5_0_defines, mulmat_body1, mulmat_load_q5_0, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_q5_0_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q5_0", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_q5_0_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q5_0", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q5_0_defines, mulmat_body1, mulmat_load_q5_0, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_q5_0_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q5_0", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_q5_0_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q5_0", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # stream.clear()
-        # stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q5_1_defines, mulmat_body1, mulmat_load_q5_1, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_q5_1_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q5_1", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_q5_1_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q5_1", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q5_1_defines, mulmat_body1, mulmat_load_q5_1, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_q5_1_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q5_1", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_q5_1_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q5_1", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # stream.clear()
-        # stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q8_0_defines, mulmat_body1, mulmat_load_q8_0, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_q8_0_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q8_0", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_q8_0_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q8_0", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q8_0_defines, mulmat_body1, mulmat_load_q8_0, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_q8_0_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q8_0", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_q8_0_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q8_0", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # stream.clear()
-        # stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q2_K_defines, mulmat_body1, mulmat_load_q2_K, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_q2_k_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q2_K", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_q2_k_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q2_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q2_K_defines, mulmat_body1, mulmat_load_q2_K, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_q2_k_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q2_K", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_q2_k_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q2_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # stream.clear()
-        # stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q3_K_defines, mulmat_body1, mulmat_load_q3_K, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_q3_k_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q3_K", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_q3_k_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q3_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q3_K_defines, mulmat_body1, mulmat_load_q3_K, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_q3_k_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q3_K", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_q3_k_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q3_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # stream.clear()
-        # stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q4_K_defines, mulmat_body1, mulmat_load_q4_K, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_q4_k_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q4_K", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_q4_k_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q4_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q4_K_defines, mulmat_body1, mulmat_load_q4_K, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_q4_k_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q4_K", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_q4_k_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q4_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # stream.clear()
-        # stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q5_K_defines, mulmat_body1, mulmat_load_q5_K, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_q5_k_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q5_K", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_q5_k_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q5_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q5_K_defines, mulmat_body1, mulmat_load_q5_K, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_q5_k_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q5_K", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_q5_k_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q5_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

-        # stream.clear()
-        # stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q6_K_defines, mulmat_body1, mulmat_load_q6_K, mulmat_body2))
-        # tasks.append(string_to_spv("matmul_id_q6_k_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q6_K", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
-        # tasks.append(string_to_spv("matmul_id_q6_k_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q6_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))
+        stream.clear()
+        stream.extend((mulmat_head, shader_int8_ext, shader_float_type, shader_q6_K_defines, mulmat_body1, mulmat_load_q6_K, mulmat_body2))
+        tasks.append(string_to_spv("matmul_id_q6_k_f32", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "A_TYPE": "block_q6_K", "B_TYPE": "float", "D_TYPE": "float"}, fp16))
+        tasks.append(string_to_spv("matmul_id_q6_k_f32_aligned", "".join(stream), {"MUL_MAT_ID": "1", "LOAD_VEC_A": 2, "LOAD_VEC_B": load_vec, "A_TYPE": "block_q6_K", "B_TYPE": vec_type, "D_TYPE": "float"}, fp16))

    # Shaders where precision is needed, so no fp16 version

@@ -3008,7 +2910,9 @@ async def main():
        stream.clear()
        stream.extend((mul_mat_vec_head, shader_int8_ext, shader_f32))

-        if i == GGML_TYPE_F16:
+        if i == GGML_TYPE_F32:
+            stream.extend((shader_f32_defines, mul_mat_vec_layout, shader_float_dequant_func, mul_mat_vec_body))
+        elif i == GGML_TYPE_F16:
            stream.extend((shader_f16_defines, mul_mat_vec_layout, shader_float_dequant_func, mul_mat_vec_body))
        elif i == GGML_TYPE_Q4_0:
            stream.extend((shader_q4_0_defines, mul_mat_vec_layout, shader_q4_0_dequant_func, mul_mat_vec_body))
@@ -3036,7 +2940,7 @@ async def main():
        tasks.append(string_to_spv(f"mul_mat_vec_{type_names[i]}_f32_f32", "".join(stream), {"B_TYPE": "float", "D_TYPE": "float", "K_QUANTS_PER_ITERATION": K_QUANTS_PER_ITERATION}))
        tasks.append(string_to_spv(f"mul_mat_vec_{type_names[i]}_f16_f32", "".join(stream), {"B_TYPE": "float16_t", "D_TYPE": "float", "K_QUANTS_PER_ITERATION": K_QUANTS_PER_ITERATION}))

-        # tasks.append(string_to_spv(f"mul_mat_vec_id_{type_names[i]}_f32", "".join(stream), {"MUL_MAT_ID": "1", "B_TYPE": "float", "D_TYPE": "float", "K_QUANTS_PER_ITERATION": K_QUANTS_PER_ITERATION}))
+        tasks.append(string_to_spv(f"mul_mat_vec_id_{type_names[i]}_f32", "".join(stream), {"MUL_MAT_ID": "1", "B_TYPE": "float", "D_TYPE": "float", "K_QUANTS_PER_ITERATION": K_QUANTS_PER_ITERATION}))

    # Dequant shaders
    for i in range(0, VK_NUM_TYPES):
@@ -3115,8 +3019,11 @@ async def main():
    tasks.append(string_to_spv("add_f32", f"{generic_binary_op_combined}\n{add_body}", {"A_TYPE": "float", "B_TYPE": "float", "D_TYPE": "float", "FLOAT_TYPE": "float"}))

    tasks.append(string_to_spv("split_k_reduce", mulmat_split_k_reduce_src, {}))
+
    tasks.append(string_to_spv("mul_f32", f"{generic_binary_op_combined}\n{mul_body}", {"A_TYPE": "float", "B_TYPE": "float", "D_TYPE": "float", "FLOAT_TYPE": "float"}))

+    tasks.append(string_to_spv("div_f32", f"{generic_binary_op_combined}\n{div_body}", {"A_TYPE": "float", "B_TYPE": "float", "D_TYPE": "float", "FLOAT_TYPE": "float"}))
+
    tasks.append(string_to_spv("scale_f32", f"{generic_unary_op_combined}\n{scale_body}", {"A_TYPE": "float", "D_TYPE": "float", "FLOAT_TYPE": "float"}))

    tasks.append(string_to_spv("sqr_f32", f"{generic_unary_op_combined}\n{sqr_body}", {"A_TYPE": "float", "D_TYPE": "float", "FLOAT_TYPE": "float"}))
@@ -3140,6 +3047,8 @@ async def main():

    tasks.append(string_to_spv("argsort_f32", argsort_src, {"A_TYPE": "float"}))

+    tasks.append(string_to_spv("sum_rows_f32", f"{generic_head}\n{shader_f32}\n{sum_rows_src}", {"A_TYPE": "float", "D_TYPE": "float"}))
+
    # Helper to decorate tasks with semaphore acquisition.
    async def withSemaphore(sem, task):
        async with sem:
--- a/gguf-py/gguf/constants.py
+++ b/gguf-py/gguf/constants.py
@@ -645,6 +645,7 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
    ],
    MODEL_ARCH.MINICPM: [
        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT,
        MODEL_TENSOR.OUTPUT_NORM,
        MODEL_TENSOR.ROPE_FREQS,
        MODEL_TENSOR.ATTN_NORM,
--- a/llama.cpp
+++ b/llama.cpp
@@ -2149,12 +2149,12 @@ struct llama_control_vector {
 struct llama_vocab {
    using id    = int32_t;
    using token = std::string;
-    using ttype = llama_token_type;
+    using tattr = llama_token_attr;

    struct token_data {
        token text;
        float score;
-        ttype type;
+        tattr attr;
    };

    enum llama_vocab_type     type     = LLAMA_VOCAB_TYPE_SPM;
@@ -2164,8 +2164,7 @@ struct llama_vocab {
    std::vector<token_data>       id_to_token;

    std::vector<id>    cache_special_tokens;
-    std::vector<token> cache_token_to_piece;         // llama_token_to_piece(special = false);
-    std::vector<token> cache_token_to_piece_special; // llama_token_to_piece(special = true);
+    std::vector<token> cache_token_to_piece; // llama_token_to_piece(special = true);

    std::map<std::pair<std::string, std::string>, int> bpe_ranks;

@@ -2372,13 +2371,34 @@ struct llama_context {
    struct llama_control_vector cvec;
 };

+static size_t llama_get_device_count(const llama_model & model) {
+    size_t count = 1;
+#if defined(GGML_USE_CUDA)
+    count = ggml_backend_cuda_get_device_count();
+#elif defined(GGML_USE_SYCL)
+    count = ggml_backend_sycl_get_device_count();
+#elif defined(GGML_USE_VULKAN)
+    count = ggml_backend_vk_get_device_count();
+#endif
+#if defined(GGML_USE_RPC)
+    count += model.rpc_servers.size();
+#endif
+    return count;
+    GGML_UNUSED(model);
+}
+
 static ggml_backend_buffer_type_t llama_default_buffer_type_offload(const llama_model & model, int gpu) {
    ggml_backend_buffer_type_t buft = nullptr;

-#ifdef GGML_USE_RPC
-    std::string endpoint = model.rpc_servers[gpu];
-    buft = ggml_backend_rpc_buffer_type(endpoint.c_str());
-#elif defined(GGML_USE_METAL)
+#if defined(GGML_USE_RPC)
+    int dev_count = (int)llama_get_device_count(model);
+    int rpc_count = (int)model.rpc_servers.size();
+    if (gpu >= dev_count - rpc_count) {
+        const char * endpoint = model.rpc_servers[gpu - dev_count + rpc_count].c_str();
+        return ggml_backend_rpc_buffer_type(endpoint);
+    }
+#endif
+#if defined(GGML_USE_METAL)
    buft = ggml_backend_metal_buffer_type();
 #elif defined(GGML_USE_CUDA)
    buft = ggml_backend_cuda_buffer_type(gpu);
@@ -2426,29 +2446,19 @@ static ggml_backend_buffer_type_t llama_default_buffer_type_split(const llama_mo
    GGML_UNUSED(tensor_split);
 }

-static size_t llama_get_device_count(const llama_model & model) {
-#if defined(GGML_USE_RPC)
-    return model.rpc_servers.size();
-#elif defined(GGML_USE_CUDA)
-    return ggml_backend_cuda_get_device_count();
-#elif defined(GGML_USE_SYCL)
-    return ggml_backend_sycl_get_device_count();
-#elif defined(GGML_USE_VULKAN)
-    return ggml_backend_vk_get_device_count();
-#else
-    return 1;
-#endif
-    GGML_UNUSED(model);
-}
-
 static size_t llama_get_device_memory(const llama_model & model, int device) {
 #if defined(GGML_USE_RPC)
-    size_t total;
-    size_t free;
-    std::string endpoint = model.rpc_servers[device];
-    ggml_backend_rpc_get_device_memory(endpoint.c_str(), &free, &total);
-    return free;
-#elif defined(GGML_USE_CUDA)
+    int dev_count = (int)llama_get_device_count(model);
+    int rpc_count = (int)model.rpc_servers.size();
+    if (device >= dev_count - rpc_count) {
+        size_t total;
+        size_t free;
+        const char * endpoint = model.rpc_servers[device - dev_count + rpc_count].c_str();
+        ggml_backend_rpc_get_device_memory(endpoint, &free, &total);
+        return free;
+    }
+#endif
+#if defined(GGML_USE_CUDA)
    size_t total;
    size_t free;
    ggml_backend_cuda_get_device_memory(device, &free, &total);
@@ -4740,7 +4750,20 @@ static void llm_load_vocab(
        auto & token_data = vocab.id_to_token[i];
        token_data.text  = std::move(word);
        token_data.score = scores ? scores[i] : 0.0f;
-        token_data.type  = toktypes ? (llama_token_type) toktypes[i] : LLAMA_TOKEN_TYPE_NORMAL;
+        token_data.attr  = LLAMA_TOKEN_ATTR_NORMAL;
+
+        if (toktypes) {  //TODO: remove, required until per token attributes are available from GGUF file
+            switch(toktypes[i]) {
+                case LLAMA_TOKEN_TYPE_UNKNOWN:      token_data.attr = LLAMA_TOKEN_ATTR_UNKNOWN;      break;
+                case LLAMA_TOKEN_TYPE_UNUSED:       token_data.attr = LLAMA_TOKEN_ATTR_UNUSED;       break;
+                case LLAMA_TOKEN_TYPE_NORMAL:       token_data.attr = LLAMA_TOKEN_ATTR_NORMAL;       break;
+                case LLAMA_TOKEN_TYPE_CONTROL:      token_data.attr = LLAMA_TOKEN_ATTR_CONTROL;      break;
+                case LLAMA_TOKEN_TYPE_USER_DEFINED: token_data.attr = LLAMA_TOKEN_ATTR_USER_DEFINED; break;
+                case LLAMA_TOKEN_TYPE_BYTE:         token_data.attr = LLAMA_TOKEN_ATTR_BYTE;         break;
+                case LLAMA_TOKEN_TYPE_UNDEFINED:    token_data.attr = LLAMA_TOKEN_ATTR_UNDEFINED;    break;
+                default:                            token_data.attr = LLAMA_TOKEN_ATTR_UNDEFINED;    break;
+            }
+        }
    }
    GGML_ASSERT(vocab.id_to_token.size() == vocab.token_to_id.size());

@@ -4831,7 +4854,7 @@ static void llm_load_vocab(
    // build special tokens cache
    {
        for (llama_vocab::id id = 0; id < (llama_vocab::id)n_vocab; ++id) {
-            if (vocab.id_to_token[id].type != LLAMA_TOKEN_TYPE_NORMAL) {
+            if (!(vocab.id_to_token[id].attr & LLAMA_TOKEN_ATTR_NORMAL)) {
                vocab.cache_special_tokens.push_back(id);
            }
        }
@@ -4845,26 +4868,75 @@ static void llm_load_vocab(
        LLAMA_LOG_INFO("%s: special tokens cache size = %u\n", __func__, (uint32_t)vocab.cache_special_tokens.size());
    }

-    // build token to piece caches
+    // build token to piece cache
    {
        size_t size_cache = 0;

-        std::vector<llama_vocab::token> cache_token_to_piece        (n_vocab);
-        std::vector<llama_vocab::token> cache_token_to_piece_special(n_vocab);
+        std::vector<llama_vocab::token> cache_token_to_piece(n_vocab);

        for (uint32_t id = 0; id < n_vocab; ++id) {
-            cache_token_to_piece[id]         = llama_token_to_piece(&model, id, false);
-            cache_token_to_piece_special[id] = llama_token_to_piece(&model, id, true);
+            cache_token_to_piece[id] = llama_token_to_piece(&model, id, true);

            size_cache += cache_token_to_piece[id].size();
-            size_cache += cache_token_to_piece_special[id].size();
        }

-        std::swap(vocab.cache_token_to_piece,         cache_token_to_piece);
-        std::swap(vocab.cache_token_to_piece_special, cache_token_to_piece_special);
+        std::swap(vocab.cache_token_to_piece, cache_token_to_piece);

        LLAMA_LOG_INFO("%s: token to piece cache size = %.4f MB\n", __func__, size_cache / 1024.0 / 1024.0);
    }
+
+    // Handle per token attributes
+    //NOTE: Each model customizes per token attributes.
+    //NOTE: Per token attributes are missing from the GGUF file.
+    //TODO: Extract attributes from GGUF file.
+    {
+        auto _contains_any = [] (const std::string &str, const std::vector<std::string> &substrs) -> bool {
+            for (auto substr : substrs) {
+                if (str.find(substr) < std::string::npos) {
+                    return true;
+                }
+            }
+            return false;
+        };
+
+        auto _set_tokenid_attr = [&] (const llama_vocab::id id, llama_token_attr attr, bool value) {
+            uint32_t current = vocab.id_to_token.at(id).attr;
+            current = value ? (current | attr) : (current & ~attr);
+            vocab.id_to_token[id].attr = (llama_token_attr) current;
+        };
+
+        auto _set_token_attr = [&] (const std::string & token, llama_token_attr attr, bool value) {
+            _set_tokenid_attr(vocab.token_to_id.at(token), attr, value);
+        };
+
+        std::string model_name;
+        std::string tokenizer_pre;
+
+        ml.get_key(LLM_KV_GENERAL_NAME, model_name, false);
+        ml.get_key(LLM_KV_TOKENIZER_PRE, tokenizer_pre, false);
+
+        // model name to lowercase
+        std::transform(model_name.begin(), model_name.end(), model_name.begin(),
+            [] (const std::string::value_type x) {
+                return std::tolower(x);
+            }
+        );
+
+        // set attributes by model/tokenizer name
+        if (_contains_any(tokenizer_pre, {"jina-v2-es", "jina-v2-de"})) {
+            _set_token_attr("<mask>", LLAMA_TOKEN_ATTR_LSTRIP, true);
+        } else if (_contains_any(model_name, {"phi-3", "phi3"})) {
+            for (auto id : vocab.cache_special_tokens) {
+                _set_tokenid_attr(id, LLAMA_TOKEN_ATTR_RSTRIP, true);
+            }
+            for (auto token : {"</s>"}) {
+                _set_token_attr(token, LLAMA_TOKEN_ATTR_RSTRIP, true);
+            }
+            for (auto token : {"<unk>", "<s>", "<|endoftext|>"}) {
+                _set_token_attr(token, LLAMA_TOKEN_ATTR_RSTRIP, false);
+            }
+        }
+    }
 }

 static void llm_load_print_meta(llama_model_loader & ml, llama_model & model) {
@@ -5129,12 +5201,10 @@ static bool llm_load_tensors(
                    // output
                    {
                        model.output_norm = ml.create_tensor(ctx_output,       tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
-                        if (model.arch != LLM_ARCH_MINICPM){
-                            model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED);
-                            // if output is NULL, init from the input tok embed
-                            if (model.output == NULL) {
-                                model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED);
-                            }
+                        model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED);
+                        // if output is NULL, init from the input tok embed
+                        if (model.output == NULL) {
+                            model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED);
                        }
                    }

@@ -10217,7 +10287,7 @@ struct llm_build_context {
        cb(cur, "lmhead_scaling", -1);

        // lm_head
-        cur = ggml_mul_mat(ctx0, model.tok_embd, cur);
+        cur = ggml_mul_mat(ctx0, model.output, cur);
        cb(cur, "result_output", -1);

        ggml_build_forward_expand(gf, cur);
@@ -12616,27 +12686,27 @@ static enum llama_vocab_type llama_vocab_get_type(const llama_vocab & vocab) {

 static bool llama_is_normal_token(const llama_vocab & vocab, llama_token id) {
    GGML_ASSERT(vocab.type != LLAMA_VOCAB_TYPE_NONE);
-    return vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_NORMAL;
+    return vocab.id_to_token[id].attr & LLAMA_TOKEN_ATTR_NORMAL;
 }

 static bool llama_is_unknown_token(const llama_vocab & vocab, llama_token id) {
    GGML_ASSERT(vocab.type != LLAMA_VOCAB_TYPE_NONE);
-    return vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_UNKNOWN;
+    return vocab.id_to_token[id].attr & LLAMA_TOKEN_ATTR_UNKNOWN;
 }

 static bool llama_is_control_token(const llama_vocab & vocab, llama_token id) {
    GGML_ASSERT(vocab.type != LLAMA_VOCAB_TYPE_NONE);
-    return vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_CONTROL;
+    return vocab.id_to_token[id].attr & LLAMA_TOKEN_ATTR_CONTROL;
 }

 static bool llama_is_byte_token(const llama_vocab & vocab, llama_token id) {
    GGML_ASSERT(vocab.type != LLAMA_VOCAB_TYPE_NONE);
-    return vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_BYTE;
+    return vocab.id_to_token[id].attr & LLAMA_TOKEN_ATTR_BYTE;
 }

 static bool llama_is_user_defined_token(const llama_vocab& vocab, llama_token id) {
    GGML_ASSERT(vocab.type != LLAMA_VOCAB_TYPE_NONE);
-    return vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_USER_DEFINED;
+    return vocab.id_to_token[id].attr & LLAMA_TOKEN_ATTR_USER_DEFINED;
 }

 static uint8_t llama_token_to_byte(const llama_vocab& vocab, llama_token id) {
@@ -13254,7 +13324,8 @@ struct fragment_buffer_variant {
 static void tokenizer_st_partition(const llama_vocab & vocab, std::forward_list<fragment_buffer_variant> & buffer) {
    // for each special token
    for (const llama_vocab::id special_id : vocab.cache_special_tokens) {
-        const auto & special_token = vocab.id_to_token[special_id].text;
+        const auto & data = vocab.id_to_token[special_id];
+        const auto & special_token = data.text;

        // for each text fragment
        std::forward_list<fragment_buffer_variant>::iterator it = buffer.begin();
@@ -13291,13 +13362,22 @@ static void tokenizer_st_partition(const llama_vocab & vocab, std::forward_list<
                    if (match > raw_text_base_offset) {
                        // left
                        const int64_t left_reminder_offset = raw_text_base_offset + 0;
-                        const int64_t left_reminder_length = match - raw_text_base_offset;
-                        buffer.emplace_after(it, raw_text, left_reminder_offset, left_reminder_length);
+                        int64_t left_reminder_length = match - raw_text_base_offset;
+
+                        if (data.attr & LLAMA_TOKEN_ATTR_LSTRIP) {
+                            while (left_reminder_length > 0 && isspace(raw_text[left_reminder_offset + left_reminder_length - 1])) {
+                                left_reminder_length--;
+                            }
+                        }
+
+                        if (left_reminder_length > 0) {
+                            buffer.emplace_after(it, raw_text, left_reminder_offset, left_reminder_length);
+                            it++;
+                        }

 #ifdef PRETOKENIZERDEBUG
                        LLAMA_LOG_WARN("FL: (%ld %ld) '%s'\n", left_reminder_offset, left_reminder_length, raw_text->substr(left_reminder_offset, left_reminder_length).c_str());
 #endif
-                        it++;
                    }

                    // special token
@@ -13306,16 +13386,25 @@ static void tokenizer_st_partition(const llama_vocab & vocab, std::forward_list<

                    // right
                    if (match + special_token.length() < raw_text_base_offset + raw_text_base_length) {
-                        const int64_t right_reminder_offset = match + special_token.length();
-                        const int64_t right_reminder_length = raw_text_base_length - ((match - raw_text_base_offset) + special_token.length());
-                        buffer.emplace_after(it, raw_text, right_reminder_offset, right_reminder_length);
+                        int64_t right_reminder_offset = match + special_token.length();
+                        int64_t right_reminder_length = raw_text_base_length - ((match - raw_text_base_offset) + special_token.length());
+
+                        if (data.attr & LLAMA_TOKEN_ATTR_RSTRIP) {
+                            while (right_reminder_length > 0 && isspace(raw_text[right_reminder_offset])) {
+                                right_reminder_offset++;
+                                right_reminder_length--;
+                            }
+                        }
+
+                        if (right_reminder_length > 0) {
+                            buffer.emplace_after(it, raw_text, right_reminder_offset, right_reminder_length);
+                            it++;
+                        }

 #ifdef PRETOKENIZERDEBUG
                        LLAMA_LOG_WARN("FR: (%ld %ld) '%s'\n", right_reminder_offset, right_reminder_length, raw_text->substr(right_reminder_offset, right_reminder_length).c_str());
 #endif

-                        it++;
-
                        if (source == 0) {
                            buffer.erase_after(buffer.before_begin());
                        } else {
@@ -13361,9 +13450,7 @@ static std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab &
                // tokenizer.encode('', add_special_tokens=True)  returns [1]
                // tokenizer.encode('', add_special_tokens=False) returns []

-                static const bool rtrim = true;  //TODO: as param
                bool is_prev_special = false;
-                bool special_token_rtrim = false;

                if (add_special && vocab.special_add_bos != 0) {
                    GGML_ASSERT(vocab.special_bos_id != -1);
@@ -13373,25 +13460,8 @@ static std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab &

                for (const auto & fragment : fragment_buffer) {
                    if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) {
-                        // without adding this leading whitespace, we do not get the same results as the original tokenizer
-
-                        // TODO: It's likely possible to get rid of this string copy entirely
-                        //  by modifying llm_tokenizer_x to operate with string offsets like pre-tokenizer
-                        //  and passing 'add space prefix' as bool argument
-                        //
                        auto raw_text = fragment.raw_text.substr(fragment.offset, fragment.length);

-                        if (special_token_rtrim) {
-                            size_t num_whitespaces = 0;
-                            while (isspace(raw_text[num_whitespaces])) {
-                                num_whitespaces++;
-                            }
-                            if (num_whitespaces == raw_text.size()) {
-                                continue; // skip if all whitespaces
-                            }
-                            raw_text = raw_text.substr(num_whitespaces);
-                        }
-
                        if (vocab.add_space_prefix) {
                            if (!output.size() || is_prev_special) {  // prefix with space if first token
                                raw_text = " " + raw_text;
@@ -13407,11 +13477,6 @@ static std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab &
                    } else { // if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN)
                        output.push_back(fragment.token);
                        is_prev_special = true;
-                        // phi-3 special tokens without rtrim, works fine for llama-spm too
-                        special_token_rtrim = rtrim
-                            && fragment.token != vocab.special_bos_id
-                            && fragment.token != vocab.special_unk_id
-                            && fragment.token != vocab.special_eos_id;
                    }
                }

@@ -14646,260 +14711,6 @@ void llama_grammar_accept_token(struct llama_context * ctx, struct llama_grammar
    ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
 }

-//
-// Beam search
-//
-
-struct llama_beam {
-    std::vector<llama_token> tokens;
-    float p;  // Cumulative beam probability (renormalized relative to all beams)
-    bool eob; // Initialize end-of-beam to false. Callback sets this to true.
-    // Sort beams by probability. In case of ties, prefer beams at eob.
-    bool operator<(const llama_beam & rhs) const {
-        return std::make_pair(p, eob) < std::make_pair(rhs.p, rhs.eob);
-    }
-    // Shift off first n tokens and discard them.
-    void shift_tokens(const size_t n) {
-        if (n) {
-            std::copy(tokens.begin() + n, tokens.end(), tokens.begin());
-            tokens.resize(tokens.size() - n);
-        }
-    }
-    llama_beam_view view() const { return {tokens.data(), tokens.size(), p, eob}; }
-};
-
-// A struct for calculating logit-related info.
-struct llama_logit_info {
-    const float * const logits;
-    const int n_vocab;
-    const float max_l;
-    const float normalizer;
-    struct sum_exp {
-        float max_l;
-        float operator()(float sum, float l) const { return sum + std::exp(l - max_l); }
-    };
-    llama_logit_info(llama_context * ctx)
-      : logits(llama_get_logits(ctx))
-      , n_vocab(llama_n_vocab(llama_get_model(ctx)))
-      , max_l(*std::max_element(logits, logits + n_vocab))
-      , normalizer(1.0f / std::accumulate(logits, logits + n_vocab, 0.0f, sum_exp{max_l}))
-      { }
-    llama_token_data get_token_data(const llama_token token_id) const {
-        constexpr auto p = std::numeric_limits<float>::quiet_NaN();  // never used
-        return {token_id, logits[token_id], p};
-    }
-    // Return top k token_data by logit.
-    std::vector<llama_token_data> top_k(size_t k) {
-        std::vector<llama_token_data> min_heap;  // min-heap by logit
-        const llama_token k_min = std::min(static_cast<llama_token>(k), n_vocab);
-        min_heap.reserve(k_min);
-        for (llama_token token_id = 0 ; token_id < k_min ; ++token_id) {
-            min_heap.push_back(get_token_data(token_id));
-        }
-        auto comp = [](const llama_token_data & a, const llama_token_data & b) { return a.logit > b.logit; };
-        std::make_heap(min_heap.begin(), min_heap.end(), comp);
-        for (llama_token token_id = k_min ; token_id < n_vocab ; ++token_id) {
-            if (min_heap.front().logit < logits[token_id]) {
-                std::pop_heap(min_heap.begin(), min_heap.end(), comp);
-                min_heap.back().id = token_id;
-                min_heap.back().logit = logits[token_id];
-                std::push_heap(min_heap.begin(), min_heap.end(), comp);
-            }
-        }
-        return min_heap;
-    }
-    float probability_from_logit(float logit) const {
-        return normalizer * std::exp(logit - max_l);
-    }
-};
-
-struct llama_beam_search_data {
-    llama_context * ctx;
-    size_t n_beams;
-    int n_past;
-    int n_predict;
-    std::vector<llama_beam> beams;
-    std::vector<llama_beam> next_beams;
-
-    // Re-calculated on each loop iteration
-    size_t common_prefix_length;
-
-    // Used to communicate to/from callback on beams state.
-    std::vector<llama_beam_view> beam_views;
-
-    llama_beam_search_data(llama_context * ctx, size_t n_beams, int n_past, int n_predict)
-      : ctx(ctx)
-      , n_beams(n_beams)
-      , n_past(n_past)
-      , n_predict(n_predict)
-      , beam_views(n_beams) {
-        beams.reserve(n_beams);
-        next_beams.reserve(n_beams);
-    }
-
-    // Collapse beams to a single beam given by index.
-    void collapse_beams(const size_t beam_idx) {
-        if (0u < beam_idx) {
-            std::swap(beams[0], beams[beam_idx]);
-        }
-        beams.resize(1);
-    }
-
-    // Min-heaps are used to efficiently collect the top-k elements (k=n_beams).
-    // The repetitive patterns below reflect the 2 stages of heaps:
-    //  * Gather elements until the vector is full, then call std::make_heap() on it.
-    //  * If the heap is full and a new element is found that should be included, pop the
-    //    least element to the back(), replace it with the new, then push it into the heap.
-    void fill_next_beams_by_top_probabilities(llama_beam & beam) {
-        // Min-heaps use a greater-than comparator.
-        const auto comp = [](const llama_beam & a, const llama_beam & b) { return a.p > b.p; };
-        if (beam.eob) {
-            // beam is at end-of-sentence, so just copy it to next_beams if its probability is high enough.
-            if (next_beams.size() < n_beams) {
-                next_beams.push_back(std::move(beam));
-                if (next_beams.size() == n_beams) {
-                    std::make_heap(next_beams.begin(), next_beams.end(), comp);
-                }
-            } else if (next_beams.front().p < beam.p) {
-                std::pop_heap(next_beams.begin(), next_beams.end(), comp);
-                next_beams.back() = std::move(beam);
-                std::push_heap(next_beams.begin(), next_beams.end(), comp);
-            }
-        } else {
-            // beam is not at end-of-sentence, so branch with next top_k tokens.
-            if (!beam.tokens.empty()) {
-                llama_decode(ctx, llama_batch_get_one(beam.tokens.data(), beam.tokens.size(), n_past, 0));
-            }
-            llama_logit_info logit_info(ctx);
-            std::vector<llama_token_data> next_tokens = logit_info.top_k(n_beams);
-
-            // Clear the kv slot so that other beams may try different tokens at this position. The llama_decode()
-            // call in loop() will conclusively fill in the kv slot once the beams converge at this position.
-            llama_kv_cache_seq_rm(ctx, 0, n_past, -1);
-
-            size_t i=0;
-            if (next_beams.size() < n_beams) {
-                for (; next_beams.size() < n_beams ; ++i) {
-                    llama_beam next_beam = beam;
-                    next_beam.tokens.push_back(next_tokens[i].id);
-                    next_beam.p *= logit_info.probability_from_logit(next_tokens[i].logit);
-                    next_beams.push_back(std::move(next_beam));
-                }
-                std::make_heap(next_beams.begin(), next_beams.end(), comp);
-            } else {
-                for (; next_beams.front().p == 0.0f ; ++i) {
-                    std::pop_heap(next_beams.begin(), next_beams.end(), comp);
-                    next_beams.back() = beam;
-                    next_beams.back().tokens.push_back(next_tokens[i].id);
-                    next_beams.back().p *= logit_info.probability_from_logit(next_tokens[i].logit);
-                    std::push_heap(next_beams.begin(), next_beams.end(), comp);
-                }
-            }
-            for (; i < n_beams ; ++i) {
-                const float next_p = beam.p * logit_info.probability_from_logit(next_tokens[i].logit);
-                if (next_beams.front().p < next_p) {
-                    std::pop_heap(next_beams.begin(), next_beams.end(), comp);
-                    next_beams.back() = beam;
-                    next_beams.back().tokens.push_back(next_tokens[i].id);
-                    next_beams.back().p = next_p;
-                    std::push_heap(next_beams.begin(), next_beams.end(), comp);
-                }
-            }
-        }
-    }
-
-    // Find common_prefix_length based on beams.
-    // Requires beams is not empty.
-    size_t find_common_prefix_length() {
-        size_t common_prefix_length = beams[0].tokens.size();
-        for (size_t i = 1 ; i < beams.size() ; ++i) {
-            common_prefix_length = std::min(common_prefix_length, beams[i].tokens.size());
-            for (size_t j = 0 ; j < common_prefix_length ; ++j) {
-                if (beams[0].tokens[j] != beams[i].tokens[j]) {
-                    common_prefix_length = j;
-                    break;
-                }
-            }
-        }
-        return common_prefix_length;
-    }
-
-    // Construct beams_state to send back to caller via the callback function.
-    // Side effect: set common_prefix_length = find_common_prefix_length();
-    llama_beams_state get_beams_state(const bool last_call) {
-        for (size_t i = 0 ; i < beams.size() ; ++i) {
-            beam_views[i] = beams[i].view();
-        }
-        common_prefix_length = find_common_prefix_length();
-        return {beam_views.data(), beams.size(), common_prefix_length, last_call};
-    }
-
-    // Loop:
-    //  * while i < n_predict, AND
-    //  * any of the beams have not yet reached end-of-beam (eob), AND
-    //  * the highest probability beam(s) (plural in case of ties) are not at end-of-sentence
-    //    (since all other beam probabilities can only decrease)
-    void loop(const llama_beam_search_callback_fn_t callback, void * const callback_data) {
-        beams.push_back({{}, 1.0f, false});  // Start with one empty beam w/ probability = 1.0 and !eob.
-        const auto not_eob = [](const llama_beam & beam) { return !beam.eob; };
-        for (int i = 0 ; i < n_predict && std::any_of(beams.begin(),beams.end(),not_eob) &&
-                       !beams[top_beam_index()].eob ; ++i) {
-            callback(callback_data, get_beams_state(false));  // Sets common_prefix_length
-            update_beams_from_beam_views();   // Update values (p,eob) that callback may have changed.
-            if (common_prefix_length) {
-                llama_decode(ctx, llama_batch_get_one(beams[0].tokens.data(), common_prefix_length, n_past, 0));
-                n_past += common_prefix_length;
-            }
-            // Zero-out next_beam probabilities to place them last in following min-heap.
-            std::for_each(next_beams.begin(), next_beams.end(), [](llama_beam & beam) { beam.p = 0.0f; });
-            for (llama_beam & beam : beams) {
-                beam.shift_tokens(common_prefix_length);
-                fill_next_beams_by_top_probabilities(beam);
-            }
-            // next_beams become the beams of next/final iteration. Swap them to re-use memory.
-            beams.swap(next_beams);
-            renormalize_beam_probabilities(beams);
-        }
-        collapse_beams(top_beam_index());
-        callback(callback_data, get_beams_state(true));
-    }
-
-    // As beams grow, the cumulative probabilities decrease.
-    // Renormalize them to avoid floating point underflow.
-    static void renormalize_beam_probabilities(std::vector<llama_beam> & beams) {
-        const auto sum_p = [](float sum, llama_beam & beam) { return sum + beam.p; };
-        const float inv_sum = 1.0f / std::accumulate(beams.begin(), beams.end(), 0.0f, sum_p);
-        std::for_each(beams.begin(), beams.end(), [=](llama_beam & beam) { beam.p *= inv_sum; });
-    }
-
-    // Assumes beams is non-empty.  Uses llama_beam::operator<() for ordering.
-    size_t top_beam_index() {
-        return std::max_element(beams.begin(), beams.end()) - beams.begin();
-    }
-
-    // Copy (p,eob) for each beam which may have been changed by the callback.
-    void update_beams_from_beam_views() {
-        for (size_t i = 0 ; i < beams.size() ; ++i) {
-            beams[i].p = beam_views[i].p;
-            beams[i].eob = beam_views[i].eob;
-        }
-    }
-};
-
-void llama_beam_search(llama_context * ctx,
-                       llama_beam_search_callback_fn_t callback, void * callback_data,
-                       size_t n_beams, int n_past, int n_predict) {
-    assert(ctx);
-    const int64_t t_start_sample_us = ggml_time_us();
-
-    llama_beam_search_data beam_search_data(ctx, n_beams, n_past, n_predict);
-
-    beam_search_data.loop(callback, callback_data);
-
-    ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
-    ctx->n_sample++;
-}
-
 //
 // quantization
 //
@@ -16167,7 +15978,7 @@ struct llama_model * llama_load_model_from_file(
            return true;
        };
    }
-    if (params.rpc_servers != nullptr) {
+    if (params.rpc_servers != nullptr && params.rpc_servers[0] != '\0') {
        // split the servers set them into model->rpc_servers
        std::string servers(params.rpc_servers);
        size_t pos = 0;
@@ -16330,17 +16141,7 @@ struct llama_context * llama_new_context_with_model(

    if (!hparams.vocab_only) {
        // initialize backends
-#if defined(GGML_USE_RPC)
-        for (auto & server : model->rpc_servers) {
-            ggml_backend_t backend = ggml_backend_rpc_init(server.c_str());
-            if (backend == nullptr) {
-                LLAMA_LOG_ERROR("%s: failed to connect RPC backend to %s\n", __func__, server.c_str());
-                llama_free(ctx);
-                return nullptr;
-            }
-            ctx->backends.push_back(backend);
-        }
-#elif defined(GGML_USE_METAL)
+#if defined(GGML_USE_METAL)
        if (model->n_gpu_layers > 0) {
            ctx->backend_metal = ggml_backend_metal_init();
            if (ctx->backend_metal == nullptr) {
@@ -16379,7 +16180,7 @@ struct llama_context * llama_new_context_with_model(
            return nullptr;
        }
        if (model->split_mode == LLAMA_SPLIT_MODE_NONE) {
-            ggml_backend_t backend = ggml_backend_vk_init(0);
+            ggml_backend_t backend = ggml_backend_vk_init(model->main_gpu);
            if (backend == nullptr) {
                LLAMA_LOG_ERROR("%s: failed to initialize Vulkan backend\n", __func__);
                llama_free(ctx);
@@ -16432,6 +16233,19 @@ struct llama_context * llama_new_context_with_model(
            }
            ctx->backends.push_back(backend);
        }
+#endif
+#if defined(GGML_USE_RPC)
+        if (model->n_gpu_layers > 0) {
+            for (const auto & endpoint : model->rpc_servers) {
+                ggml_backend_t backend = ggml_backend_rpc_init(endpoint.c_str());
+                if (backend == nullptr) {
+                    LLAMA_LOG_ERROR("%s: failed to initialize RPC to '%s'\n", __func__, endpoint.c_str());
+                    llama_free(ctx);
+                    return nullptr;
+                }
+                ctx->backends.push_back(backend);
+            }
+        }
 #endif
        ctx->backend_cpu = ggml_backend_cpu_init();
        if (ctx->backend_cpu == nullptr) {
@@ -18214,9 +18028,9 @@ float llama_token_get_score(const struct llama_model * model, llama_token token)
    return model->vocab.id_to_token[token].score;
 }

-llama_token_type llama_token_get_type(const struct llama_model * model, llama_token token) {
+llama_token_attr llama_token_get_attr(const struct llama_model * model, llama_token token) {
    GGML_ASSERT(model->vocab.type != LLAMA_VOCAB_TYPE_NONE);
-    return model->vocab.id_to_token[token].type;
+    return model->vocab.id_to_token[token].attr;
 }

 bool llama_token_is_eog(const struct llama_model * model, llama_token token) {
@@ -18318,9 +18132,14 @@ static std::string llama_decode_text(const std::string & text) {

 // does not write null-terminator to buf
 int32_t llama_token_to_piece(const struct llama_model * model, llama_token token, char * buf, int32_t length, bool special) {
+    // ref: https://github.com/ggerganov/llama.cpp/pull/7587#discussion_r1620983843
+    if (!special && llama_is_control_token(model->vocab, token)) {
+        return 0;
+    }
+
    // if we have a cache - use it
    {
-        const auto & cache = special ? model->vocab.cache_token_to_piece_special : model->vocab.cache_token_to_piece;
+        const auto & cache = model->vocab.cache_token_to_piece;

        if (!cache.empty()) {
            const auto & res = cache.at(token);
--- a/llama.h
+++ b/llama.h
@@ -97,7 +97,7 @@ extern "C" {
        LLAMA_ROPE_TYPE_GLM  =  4,
    };

-    enum llama_token_type {
+    enum llama_token_type { //TODO: remove, required until per token attributes are available from GGUF file
        LLAMA_TOKEN_TYPE_UNDEFINED    = 0,
        LLAMA_TOKEN_TYPE_NORMAL       = 1,
        LLAMA_TOKEN_TYPE_UNKNOWN      = 2,
@@ -107,6 +107,20 @@ extern "C" {
        LLAMA_TOKEN_TYPE_BYTE         = 6,
    };

+    enum llama_token_attr {
+        LLAMA_TOKEN_ATTR_UNDEFINED    = 0,
+        LLAMA_TOKEN_ATTR_UNKNOWN      = 1 <<  1,
+        LLAMA_TOKEN_ATTR_UNUSED       = 1 <<  2,
+        LLAMA_TOKEN_ATTR_NORMAL       = 1 <<  3,
+        LLAMA_TOKEN_ATTR_CONTROL      = 1 <<  4,  // SPECIAL?
+        LLAMA_TOKEN_ATTR_USER_DEFINED = 1 <<  5,
+        LLAMA_TOKEN_ATTR_BYTE         = 1 <<  6,
+        LLAMA_TOKEN_ATTR_NORMALIZED   = 1 <<  7,
+        LLAMA_TOKEN_ATTR_LSTRIP       = 1 <<  8,
+        LLAMA_TOKEN_ATTR_RSTRIP       = 1 <<  9,
+        LLAMA_TOKEN_ATTR_SINGLE_WORD  = 1 << 10,
+    };
+
    // model file types
    enum llama_ftype {
        LLAMA_FTYPE_ALL_F32              = 0,
@@ -821,7 +835,7 @@ extern "C" {

    LLAMA_API float llama_token_get_score(const struct llama_model * model, llama_token token);

-    LLAMA_API enum llama_token_type llama_token_get_type(const struct llama_model * model, llama_token token);
+    LLAMA_API enum llama_token_attr llama_token_get_attr(const struct llama_model * model, llama_token token);

    // Check if the token is supposed to end generation (end-of-generation, eg. EOS, EOT, etc.)
    LLAMA_API bool llama_token_is_eog(const struct llama_model * model, llama_token token);
@@ -1042,49 +1056,9 @@ extern "C" {
                     llama_token   token);

    //
-    // Beam search
+    // Model split
    //

-    struct llama_beam_view {
-        const llama_token * tokens;
-
-        size_t n_tokens;
-        float  p;        // Cumulative beam probability (renormalized relative to all beams)
-        bool   eob;      // Callback should set this to true when a beam is at end-of-beam.
-    };
-
-    // Passed to beam_search_callback function.
-    // Whenever 0 < common_prefix_length, this number of tokens should be copied from any of the beams
-    // (e.g. beams[0]) as they will be removed (shifted) from all beams in all subsequent callbacks.
-    // These pointers are valid only during the synchronous callback, so should not be saved.
-    struct llama_beams_state {
-        struct llama_beam_view * beam_views;
-
-        size_t n_beams;               // Number of elements in beam_views[].
-        size_t common_prefix_length;  // Current max length of prefix tokens shared by all beams.
-        bool   last_call;             // True iff this is the last callback invocation.
-    };
-
-    // Type of pointer to the beam_search_callback function.
-    // void* callback_data is any custom data passed to llama_beam_search, that is subsequently
-    // passed back to beam_search_callback. This avoids having to use global variables in the callback.
-    typedef void (*llama_beam_search_callback_fn_t)(void * callback_data, struct llama_beams_state);
-
-    /// @details Deterministically returns entire sentence constructed by a beam search.
-    /// @param ctx Pointer to the llama_context.
-    /// @param callback Invoked for each iteration of the beam_search loop, passing in beams_state.
-    /// @param callback_data A pointer that is simply passed back to callback.
-    /// @param n_beams Number of beams to use.
-    /// @param n_past Number of tokens already evaluated.
-    /// @param n_predict Maximum number of tokens to predict. EOS may occur earlier.
-    LLAMA_API void llama_beam_search(
-                   struct llama_context * ctx,
-        llama_beam_search_callback_fn_t   callback,
-                                   void * callback_data,
-                                 size_t   n_beams,
-                                int32_t   n_past,
-                                int32_t   n_predict);
-
    /// @details Build a split GGUF final path for this chunk.
    ///          llama_split_path(split_path, sizeof(split_path), "/models/ggml-model-q4_0", 2, 4) => split_path = "/models/ggml-model-q4_0-00002-of-00004.gguf"
    //  Returns the split_path length.
--- a/models/ggml-vocab-phi-3.gguf
+++ b/models/ggml-vocab-phi-3.gguf
--- a/scripts/compare-commits.sh
+++ b/scripts/compare-commits.sh
@@ -10,16 +10,18 @@ set -x

 bench_args="${@:3}"

-rm -f llama-bench.sqlite
+rm -f llama-bench.sqlite > /dev/null

 # to test a backend, call the script with the corresponding environment variable (e.g. LLAMA_CUDA=1 ./scripts/compare-commits.sh ...)

-git checkout $1
-make clean && make -j32 $make_opts llama-bench
-./llama-bench -o sql $bench_args | tee /dev/tty | sqlite3 llama-bench.sqlite
+git checkout $1 > /dev/null
+make clean > /dev/null
+make -j$(nproc) $make_opts llama-bench > /dev/null
+./llama-bench -o sql -oe md $bench_args | sqlite3 llama-bench.sqlite

-git checkout $2
-make clean && make -j32 $make_opts llama-bench
-./llama-bench -o sql $bench_args | tee /dev/tty | sqlite3 llama-bench.sqlite
+git checkout $2 > /dev/null
+make clean > /dev/null
+make -j$(nproc) $make_opts llama-bench > /dev/null
+./llama-bench -o sql -oe md $bench_args | sqlite3 llama-bench.sqlite

 ./scripts/compare-llama-bench.py -b $1 -c $2
--- a/tests/test-tokenizer-random.py
+++ b/tests/test-tokenizer-random.py
@@ -156,17 +156,39 @@ def generator_custom_text_edge_cases() -> Iterator[str]:
        '<s>a',       # Phi-3 fail
        '<unk><|endoftext|><s>',  # Phi-3 fail
        'a\na',       # TODO: Bert fail
+        'a </s> b',   # rstrip phi-3
+        'a <mask> b', # lstrip jina-v2
    ]


-def generator_random_special_tokens(tokenizer, iterations=100) -> Iterator[str]:
-    special_tokens = set(tokenizer.all_special_tokens)
-    special_tokens.update([" ", "\n", "\t", "-", "!", "one", "1", "<s>", "</s>"])
-    special_tokens = list(sorted(special_tokens))
+def generator_vocab_words(vocab: list[str]) -> Iterator[str]:
+    """Brute force check all vocab words"""
+    yield from vocab
+
+
+def generator_added_lr_strip(tokenizer) -> Iterator[str]:
+    WHITESPACES = ["", " ", "  ", "    "]
+    special_tokens = list(tokenizer.all_special_tokens)
+    added_tokens   = list(tokenizer.added_tokens_encoder)
+    all_tokens     = list(sorted(set(special_tokens + added_tokens)))
+    for token in all_tokens:
+        for lstrip in WHITESPACES:
+            for rstrip in WHITESPACES:
+                yield lstrip + token + rstrip
+                yield "a" + lstrip + token + rstrip
+                yield lstrip + token + rstrip + "z"
+                yield "a" + lstrip + token + rstrip + "z"
+
+
+def generator_random_added_tokens(tokenizer, iterations=100) -> Iterator[str]:
+    special_tokens = list(tokenizer.all_special_tokens)
+    added_tokens   = list(tokenizer.added_tokens_encoder)
+    separations    = [" ", "\n", "\t", "-", "!", "one", "1", "<s>", "</s>"]
+    all_tokens     = list(sorted(set(special_tokens + added_tokens + separations)))
    rand = random.Random()
    for m in range(iterations):
        rand.seed(m)
-        words = rand.choices(special_tokens, k=500)
+        words = rand.choices(all_tokens, k=500)
        if words[0] == tokenizer.bos_token:  # skip spam warning of double BOS
            while len(words) > 1 and words[1] == tokenizer.bos_token:  # leave one starting BOS
                words.pop(0)
@@ -175,11 +197,6 @@ def generator_random_special_tokens(tokenizer, iterations=100) -> Iterator[str]:
        yield "".join(words)


-def generator_vocab_words(vocab: list[str]) -> Iterator[str]:
-    """Brute force check all vocab words"""
-    yield from vocab
-
-
 def generator_random_chars(iterations=100) -> Iterator[str]:
    """Brute force random text with simple characters"""

@@ -274,8 +291,8 @@ def test_compare_tokenizer(func_tokenize1: Callable, func_tokenize2: Callable, g
        ids2 = func_tokenize2(text)
        if ids1 != ids2:
            i = find_first_mismatch(ids1, ids2)
-            ids1 = list(ids1)[max(0, i - 2) : i + 2 + 1]
-            ids2 = list(ids2)[max(0, i - 2) : i + 2 + 1]
+            ids1 = list(ids1)[max(0, i - 2) : i + 5 + 1]
+            ids2 = list(ids2)[max(0, i - 2) : i + 5 + 1]
            logger.info(" TokenIDs: " + str(ids1))
            logger.info(" Expected: " + str(ids2))
            raise Exception()
@@ -309,8 +326,9 @@ def main(argv: list[str] = None):
    vocab = list(sorted(tokenizer.batch_decode(list(tokenizer.get_vocab().values()), skip_special_tokens=True)))
    test_compare_tokenizer(func_tokenize1, func_tokenize2, generator_custom_text())
    test_compare_tokenizer(func_tokenize1, func_tokenize2, generator_custom_text_edge_cases())
-    test_compare_tokenizer(func_tokenize1, func_tokenize2, generator_random_special_tokens(tokenizer, 10_000))
    test_compare_tokenizer(func_tokenize1, func_tokenize2, generator_vocab_words(vocab))
+    test_compare_tokenizer(func_tokenize1, func_tokenize2, generator_added_lr_strip(tokenizer))
+    test_compare_tokenizer(func_tokenize1, func_tokenize2, generator_random_added_tokens(tokenizer, 10_000))
    test_compare_tokenizer(func_tokenize1, func_tokenize2, generator_random_chars(10_000))
    test_compare_tokenizer(func_tokenize1, func_tokenize2, generator_random_vocab_chars(vocab, 10_000))
    test_compare_tokenizer(func_tokenize1, func_tokenize2, generator_random_vocab_words(vocab, 5_000))
@@ -322,14 +340,14 @@ def main(argv: list[str] = None):
 if __name__ == "__main__":
    # main()

-    path_tokenizers = "./models/tokenizers/"
+    path_tokenizers   = "./models/tokenizers/"
    path_vocab_format = "./models/ggml-vocab-%s.gguf"

    # import os
    # tokenizers = os.listdir(path_tokenizers)
    tokenizers = [
-        # "llama-spm",   # SPM
-        # "phi-3",       # SPM
+        "llama-spm",   # SPM
+        "phi-3",       # SPM
        "jina-v2-en",  # WPM
        "bert-bge",    # WPM
    ]
Author	SHA1	Message	Date
Georgi Gerganov	0cd6bd3483	llama : remove beam search (#7736 )	2024-06-04 21:23:05 +03:00
Georgi Gerganov	5ca0944a15	readme : remove obsolete Zig instructions (#7471 )	2024-06-04 19:43:01 +03:00
slaren	adc9ff3841	llama-bench : allow using a different printer for stderr with -oe (#7722 ) compare-commits.sh : hide stdout, use -oe to print markdown	2024-06-04 14:32:42 +02:00
Daniele	987d743d6b	Improve hipBLAS support in CMake (#7696 ) * Improve hipBLAS support in CMake This improves the detection of the correct CMAKE_PREFIX_PATH when using different distributions or a self-built ROCm SDK. * Set ROCM_PATH correctly	2024-06-04 14:09:15 +02:00
zhouwg	b226c1227b	refine .gitignore (#7688 ) This adds tags and android ndk into the git ignore list	2024-06-04 21:21:26 +10:00
jaime-m-p	3b38d48609	Per token attributes (#7685 ) * Add per token attributes enum * Using phi-3 for testing 'rstrip' * Using jina-v2 for testing 'lstrip' * Brute force test for 'lstrip' and 'rstrip' * Implement 'rstrip' and 'lstrip' * Update phi-3 GGUF file (obsolete since `917dc8c`) * Replace llama_token_type with llama_token_attribs	2024-06-04 09:17:17 +02:00
Georgi Gerganov	6d1616944d	ggml : prevent builds with -ffinite-math-only (#7726 ) This enforces a check that -fno-finite-math-only was set and that the operating compiling mode is not in finite maths mode. This is because during rewriting of silu and softmax for cpu #7154 there emerged an issue where the result that was observed when >1 slot was nondeterministic as found by @JohannesGaessler. @LostRuins narrowed the problem down to -ffinite-math-only which was theorised to be due to SiLU, instead of flushing small values to 0, returns NaN or some other garbage. @jart proposed a fix that @ggerganov then implemented in this fix ref https://github.com/ggerganov/llama.cpp/pull/7154#issuecomment-2145661825	2024-06-04 17:01:09 +10:00
Radoslav Gerganov	bde7cd3cd9	llama : offload to RPC in addition to other backends (#7640 ) * llama : offload to RPC in addition to other backends * - fix copy_tensor being called on the src buffer instead of the dst buffer - always initialize views in the view_src buffer - add RPC backend to Makefile build - add endpoint to all RPC object names * add rpc-server to Makefile * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>	2024-06-03 20:03:26 +03:00
Masaya, Kato	a5735e4426	ggml : use OpenMP as a thread pool (#7606 ) * ggml: Added OpenMP for multi-threads processing * ggml : Limit the number of threads used to avoid deadlock * update shared state n_threads in parallel region * clear numa affinity for main thread even with openmp * enable openmp by default * fix msvc build * disable openmp on macos * ci : disable openmp with thread sanitizer * Update ggml.c Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>	2024-06-03 17:14:15 +02:00
Johannes Gäßler	0b832d53ba	make: fix debug options not being applied to NVCC (#7714 )	2024-06-03 16:28:58 +02:00
0cc4m	3d7ebf6312	Vulkan Mixture of Experts (MoE) support (#7628 ) * Finish Vulkan mul_mat_id implementation * Add Vulkan sum_rows and div ops * Fix MUL_MAT_ID matrix matrix shader * Fix MUL_MAT_ID matrix vector shader dispatch size * Fix MUL_MAT_ID matrix vector shader and dispatch code * Update Vulkan CPU offload for MUL_MAT_ID * Fix crash when using split mode none and setting a main GPU	2024-06-03 10:59:14 +02:00
Andy Tai	a10cda58d3	cmake : add pkg-config spec file for llama.cpp (#7702 )	2024-06-03 11:06:24 +03:00
zhangkaihuo	6f28a333c1	llama : MiniCPM support tied embeddings (#7664 ) * support lm_head * remove the code block --------- Co-authored-by: zhangkaihuo <zhangkaihuo@modelbest.cn>	2024-06-03 10:49:30 +03:00
Georgi Gerganov	549279d804	llama : avoid double token-to-piece cache (#7654 ) ggml-ci	2024-06-03 08:34:43 +03:00