ngxson/llama.cpp

Pull Requests Commits

llama : default pooling last for qwen3

ngxson committed 198 days ago

23c5b57a

llama : allow using mmap without PrefetchVirtualMemory, apply GGML_WIN_VER to llama.cpp sources (#14013)

slaren committed 198 days ago

Verified 3a077146

readme : add badge (#13938)

Olexandr88 committed 198 days ago

Verified d01d112a

vocab : warn about missing mask token (#14022)

CISC committed 198 days ago

Verified 9f47fa57

context : fix pos_min initialization upon error decode (#14008)

ggerganov committed 198 days ago

Verified 9e31bec4

vulkan: automatically deduce size of push constants (#13936)

jeffbolznv committed 198 days ago

Verified 5a8ae305

ggml-vulkan: adds support for op CONV_TRANSPOSE_1D (#13813)

etasnadi committed 199 days ago

Verified 0d398442

kv-cache : refactor the update/defrag mechanism (#13988)

ggerganov committed 199 days ago

Verified 3e63a58e

ci : remove cuda 11.7 releases, switch runner to windows 2022 (#13997)

slaren committed 199 days ago

Verified 2589ad37

releases : use dl backend for linux release, remove arm64 linux release (#13996)

slaren committed 199 days ago

Verified 48254871

llama-graph : use ggml_repeat_4d (#13998)

ngxson committed 199 days ago

Verified 3ac67535

CUDA: fix FTZ in FA for Gemma 3 (#13991)

JohannesGaessler committed 199 days ago

Verified 0b4be4c4

kv-cache : fix unified::seq_rm to work with seq_id < 0 (#13985)

ggerganov committed 199 days ago

Verified e0e806f5

vulkan: fix warnings in perf logger querypool code (#13937)

jeffbolznv committed 200 days ago

Verified 7e00e60e

docs : add "Quick start" section for new users (#13862)

ngxson committed 200 days ago

Verified ea1431b0

opencl: add `backend_synchronize` (#13939)

lhez committed 200 days ago

Verified 71e74a3a

OpenCL: Add concat, tsembd, upscale, tanh, pad and repeat (#13840)

rmatif committed 200 days ago

Verified bfb1e012

server : disable speculative decoding for SWA models (#13970)

ggerganov committed 201 days ago

Verified 36375762

metal : use F32 accumulators in FA kernels (#13975)

ggerganov committed 201 days ago

Verified ea394d7a

gemma : more consistent attention scaling for v2 and v3 (#13951)

ggerganov committed 201 days ago

Verified 5582c49c

`server`: update deepseek reasoning format (pass reasoning_content as diffs) (#13933)

ochafik committed 201 days ago

Verified c9bbc779

mtmd : fix memory leak in mtmd_helper_eval_chunk_single (#13961)

ngxson committed 201 days ago

Verified bfd32279

cmake : Handle mixed-case 'Power' strings in POWER CPU detection (#13966)

shalinib-ibm committed 201 days ago

Verified 093e3f1f

sycl: quantize and reorder the input to q8_1 when reorder is enabled (#13826)

AD2605 committed 201 days ago

Verified 663445b0

gguf: fix failure on version == 0 (#13956)

JohannesGaessler committed 202 days ago

Verified 7675c555

convert : fix nomic-bert-moe mask token (#13757)

CISC committed 202 days ago

Verified 5e1c3aed

convert : fix vocab padding code for bert models (#13954)

CISC committed 202 days ago

Verified c496fe0b

ggml: check if non-native endian model is being loaded (#13943)

taronaeo committed 202 days ago

Verified e57bb87c

ggerganov committed 202 days ago

f3a4b165

vulkan : Remove unexpected ; (ggml/1253)

dg0yt committed 202 days ago

108009f5

Older