ggerganov/llama.cpp

Pull Requests Commits

vulkan: Implement grouped query attention in the coopmat2 FA shader (#12559)

jeffbolznv committed 322 days ago

Verified be0a0f8c

Vulkan: Fix mmq int dot float cache size (#12722)

0cc4m committed 322 days ago

Verified 92e3006b

model : print tensor size during load (#12711)

ggerganov committed 322 days ago

Verified 833e2b74

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

slaren committed 322 days ago

Verified e0e912f4

llama : refactor kv cache guard (#12695)

ggerganov committed 323 days ago

Verified a10b36c9

vocab : BailingMoE : change possessive quantifiers to greedy (#12677)

CISC committed 323 days ago

Verified 83a88bd6

common : remove json.hpp from common.cpp (#12697)

ngxson committed 323 days ago

Verified 42eb248f

[CANN] get_rows and dup optimization (#12671)

noemotiovon committed 323 days ago

Verified 9bacd6b3

common : refactor downloading system, handle mmproj with -hf option (#12694)

ngxson committed 323 days ago

Verified 267c1399

opencl : fix memory allocation size (#12649)

sparkleholic committed 323 days ago

Verified f423981a

llama : use LLM_KV_GENERAL_FILE_TYPE instead of gguf_find_key (#12672)

jklincn committed 323 days ago

Verified e39e727e

convert : BailingMoE : fix qkv split when head_dim is 0 (#12687)

CISC committed 323 days ago

Verified 5936a616

metal : use F32 prec in FA kernels (#12688)

ggerganov committed 323 days ago

Verified 3fd072a5

Fix clang warning in gguf_check_reserved_keys (#12686)

yeahdongcn committed 324 days ago

Verified a6f32f0b

vulkan: fix build when glslc doesn't support coopmat (#12683)

wbruna committed 324 days ago

Verified 2bb3597e

SYCL: Rename oneMKL to oneMath (#12192)

Rbiessy committed 324 days ago

Verified 82939705

SYCL: switch to SYCL namespace (#12674)

qnixsynapse committed 324 days ago

Verified 8bbf2608

convert : BailingMoE : avoid setting rope_dim to 0 (#12678)

CISC committed 324 days ago

Verified 35782aee

vocab : add special infill tokens for CodeLlama (#11850)

danbev committed 324 days ago

Verified c80a7759

ggml : faster ssm scan (#10558)

A3shTnT committed 324 days ago

Verified 250d7953

convert : Qwerky : use lora_rank_tokenshift and lora_rank_decay if present (#12667)

CISC committed 324 days ago

Verified 403fbacb

Vulkan: Add DP4A MMQ and Q8_1 quantization shader (#12135)

0cc4m committed 324 days ago

Verified a8a1f335

cmake : fix whitespace (#0)

ggerganov committed 324 days ago

1790e731

ggerganov committed 324 days ago

0114a32d

cmake: improve Vulkan cooperative matrix support checks (whisper/2966)

sandrohanea committed 324 days ago

a7724480

llava : proper description fix (#12668)

CISC committed 325 days ago

Verified 1a859490

SYCL: Remove misleading ggml_sycl_op_flatten function (#12387)

qnixsynapse committed 325 days ago

Verified 6c02a032

llava : fix clip loading GGUFs with missing description (#12660)

CISC committed 325 days ago

Verified f52d59d7

tts : remove printfs (#12640)

marcoStocchi committed 325 days ago

Verified 52de2e59

llama : support BailingMoE (Ling) (#12634)

CISC committed 325 days ago

Verified 2c3f8b85

Newer Older