ngxson/llama.cpp

Pull Requests Commits

HIP: fix flash_attn_stream_k_fixup warning (#11604)

JohannesGaessler committed 1 year ago

Verified 6eecde3c

CUDA/HIP: add support for selectable warp size to mmv (#11519)

IMbackK committed 1 year ago

Verified 396856b4

HIP: add GGML_CUDA_CC_IS_* for amd familys as increasing cc archtectures for amd gpus are not supersets of eatch other (#11601)

IMbackK committed 1 year ago

Verified 4d0598e1

nit: more informative crash when grammar sampler fails (#11593)

ochafik committed 1 year ago

Verified 90f9b88a

CUDA: use mma PTX instructions for FlashAttention (#11583)

JohannesGaessler committed 1 year ago

Verified 864a0b67

Name colors (#11573)

ericcurtin committed 1 year ago

Verified 84ec8a58

`tool-call`: support Command R7B (+ return tool_plan "thoughts" in API) (#11585)

ochafik committed 1 year ago

Verified bfcce4d6

Fix exotic ci env that lacks ostringstream::str (#11581)

ochafik committed 1 year ago

Verified 69804487

sampling : support for llguidance grammars (#10224)

mmoskal committed 1 year ago

Verified ff227703

llama : add support for GLM-Edge and GLM-Edge-V series models (#10573)

piDack committed 1 year ago

Verified 0cec062a

ci: use sccache on windows HIP jobs (#11553)

ochafik committed 1 year ago

Verified 53debe6f

`sync`: minja (https://github.com/google/minja/commit/418a2364b56dc9be4ed9a1a2b0fb16fb53a7a22e) (#11574)

ochafik committed 1 year ago

Verified cfd74c86

Implement s3:// protocol (#11511)

ericcurtin committed 1 year ago

Verified ecef206c

ci: simplify cmake build commands (#11548)

ochafik committed 1 year ago

Verified 5bbc7362

`ci`: use sccache on windows instead of ccache (#11545)

ochafik committed 1 year ago

Verified aa6fb132

`tool-call`: fix llama 3.x and functionary 3.2, play nice w/ pydantic_ai package, update readme (#11539)

ochafik committed 1 year ago

Verified a83f5286

fix stop regression (#11543)

ochafik committed 1 year ago

Verified b1bcd309

Fix chatml fallback for unsupported builtin templates (when --jinja not enabled) (#11533)

ochafik committed 1 year ago

Verified 5783575c

server : fix --jinja when there's no tools or schema (typo was forcing JSON) (#11531)

ochafik committed 1 year ago

Verified 4a2b196d

common: Add missing va_end (#11529)

stevegrubb committed 1 year ago

Verified 1bd3047a

server : update help metrics processing/deferred (#11512)

danbev committed 1 year ago

Verified a2df2787

`ci`: ccache for all github worfklows (#11516)

ochafik committed 1 year ago

Verified 553f1e46

Tool call support (generic + native for Llama, Functionary, Hermes, Mistral, Firefunction, DeepSeek) w/ lazy grammars (#9639)

ochafik committed 1 year ago

Verified 8b576b6c

HIP: require at least HIP 5.5

IMbackK committed 1 year ago

27d135c9

HIP: Prepare reduction operators for wave 64

IMbackK committed 1 year ago

6af1ca48

CUDA/HIP: add warp_size to cuda_device_info

IMbackK committed 1 year ago

c300e68e

sync: minja (#11499)

ochafik committed 1 year ago

Verified 3d804dec

vocab : correctly identify LF token for GPT-2 style BPE tokenizer (#11496)

mgroeber9110 committed 1 year ago

Verified ffd0821c

server : use lambda instead of std::bind (#11507)

danbev committed 1 year ago

Verified 4314e56c

server : (docs) added response format for /apply-template [no ci] (#11503)

isaac-mcfadyen committed 1 year ago

Verified 496e5bf4

Newer Older