Georgi Gerganov e298d2fbd0 kv-cache : add SWA support (#13194) ...

* kv-cache : prepare for SWA

ggml-ci

* kv-cache : initial iSWA implementation

ggml-ci

* kv-cache : rework error recovery logic

ggml-ci

* models : fix Phi-3 SWA parameters

ggml-ci

* model : adjust Granite to rope factor changes

ggml-ci

* server : check if context can do shifts

ggml-ci

* iswa : for now, always enable shifts (experiment)

ggml-ci

* kv-cache : simplify SWA logic

ggml-ci

* kv-cache : apply defrag when we fail to find slots for the batch

ggml-ci

* llama : update docs about llama_decode

ggml-ci

* kv-cache : update warning logs when no space for the batch is available

ggml-ci

* llama : add llama_kv_self_seq_pos_min()

* kv-cache : keep track of partial SWA computes and print warnings

* server : disallow use cases involving partial SWA context

ggml-ci

* llama : add param to control SWA cache size

ggml-ci

* minor : clean-up

ggml-ci

2025-05-20 08:05:46 +03:00

..

CMakeLists.txt

llama/ggml: add LLM training support (#10544)

2025-05-12 14:44:49 +02:00

llama-adapter.cpp

llama : do not crash if there is no CPU backend (#13395)

2025-05-09 13:02:07 +02:00

llama-adapter.h

llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)

2025-03-13 12:35:44 +02:00

llama-arch.cpp

model : Granite MoE shared (#13269)

2025-05-13 15:12:01 +02:00

llama-arch.h

model : Nomic Embed Text V2 with Mixture-of-Experts (MoE) architecture (#12466)

2025-04-28 22:52:15 +03:00

llama-batch.cpp

kv-cache : separate recurrent vs non-recurrent impl (#12799)

2025-05-02 17:48:36 +03:00

llama-batch.h

kv-cache : separate recurrent vs non-recurrent impl (#12799)

2025-05-02 17:48:36 +03:00

llama-chat.cpp

llama : one-off chat template fix for Mistral-Small-2503 (#13398)

2025-05-09 11:17:51 +02:00

llama-chat.h

llama : one-off chat template fix for Mistral-Small-2503 (#13398)

2025-05-09 11:17:51 +02:00

llama-context.cpp

kv-cache : add SWA support (#13194)

2025-05-20 08:05:46 +03:00

llama-context.h

llama/ggml: add LLM training support (#10544)

2025-05-12 14:44:49 +02:00

llama-cparams.cpp

llama : refactor src/llama.cpp (#10902)

2025-01-03 10:18:53 +02:00

llama-cparams.h

Add --no-op-offload to improve -ot pp perf in MoE models like llama4 400B (#13386)

2025-05-11 14:18:39 +02:00

llama-grammar.cpp

tool-call: fix Qwen 2.5 Coder support, add micro benchmarks, support trigger patterns for lazy grammars (#12034)

2025-03-05 13:05:13 +00:00

llama-grammar.h

tool-call: fix Qwen 2.5 Coder support, add micro benchmarks, support trigger patterns for lazy grammars (#12034)

2025-03-05 13:05:13 +00:00

llama-graph.cpp

kv-cache : add SWA support (#13194)

2025-05-20 08:05:46 +03:00

llama-graph.h

kv-cache : add SWA support (#13194)

2025-05-20 08:05:46 +03:00

llama-hparams.cpp

hparams : add SWA rope parameters (#12374)

2025-03-14 09:03:24 +02:00

llama-hparams.h

kv-cache : add SWA support (#13194)

2025-05-20 08:05:46 +03:00

llama-impl.cpp

GGUF: C++ refactor, backend support, misc fixes (#11030)

2025-01-07 18:01:58 +01:00

llama-impl.h

cleanup: fix compile warnings associated with gnu_printf (#11811)

2025-02-12 10:06:53 -04:00

llama-io.cpp

llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)

2025-03-13 12:35:44 +02:00

llama-io.h

llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)

2025-03-13 12:35:44 +02:00

llama-kv-cache.cpp

kv-cache : add SWA support (#13194)

2025-05-20 08:05:46 +03:00

llama-kv-cache.h

kv-cache : add SWA support (#13194)

2025-05-20 08:05:46 +03:00

llama-memory.cpp

llama : refactor llama_context, llama_kv_cache, llm_build_context (#12181)

2025-03-13 12:35:44 +02:00

llama-memory.h

kv-cache : add SWA support (#13194)

2025-05-20 08:05:46 +03:00

llama-mmap.cpp

mmap : skip resource limit checks on AIX (#12541)

2025-03-24 12:17:10 +02:00

llama-mmap.h

llama-mmap: fix missing include (#11796)

2025-02-10 20:58:18 +02:00

llama-model-loader.cpp

gguf : use ggml log system (#13571)

2025-05-15 19:13:11 +02:00

llama-model-loader.h

llama : add option to override model tensor buffers (#11397)

2025-04-02 14:52:01 +02:00

llama-model-saver.cpp

llama/ggml: add LLM training support (#10544)

2025-05-12 14:44:49 +02:00

llama-model-saver.h

llama/ggml: add LLM training support (#10544)

2025-05-12 14:44:49 +02:00

llama-model.cpp

kv-cache : add SWA support (#13194)

2025-05-20 08:05:46 +03:00

llama-model.h

kv-cache : add SWA support (#13194)

2025-05-20 08:05:46 +03:00

llama-quant.cpp

quantize : improve tensor-type pattern matching (#13033)

2025-05-13 19:12:31 +02:00

llama-quant.h

llama : refactor src/llama.cpp (#10902)

2025-01-03 10:18:53 +02:00

llama-sampling.cpp

sampling : make top_n_sigma no-op at <=0 or a single candidate (#13345)

2025-05-06 22:36:24 +02:00

llama-sampling.h

llama : add llama_vocab, functions -> methods, naming (#11110)

2025-01-12 11:32:42 +02:00

llama-vocab.cpp

llama/ggml: add LLM training support (#10544)

2025-05-12 14:44:49 +02:00

llama-vocab.h

llama/ggml: add LLM training support (#10544)

2025-05-12 14:44:49 +02:00

llama.cpp

llama : print hint when loading a model when no backends are loaded (#13589)

2025-05-16 16:38:07 +02:00

unicode-data.cpp

server : better security control for public deployments (#9776)

2024-10-08 13:27:04 +02:00

unicode-data.h

llama : reduce compile time and binary size (#9712)

2024-10-02 15:49:55 +02:00

unicode.cpp

repo : update links to new url (#11886)

2025-02-15 16:40:57 +02:00

unicode.h

unicode : improve naming style (#10838)

2024-12-16 12:31:45 +02:00