ggerganov/llama.cpp

Pull Requests Commits

graph : make llm_graph_context destructor virtual (#14410)

ggerganov committed 1 year ago

Verified 72babea5

recurrent : call balloc split_reset() in init_batch() (#14414)

ggerganov committed 1 year ago

Verified 43678060

ggml : add ggml_set_rows (#14274)

rgerganov committed 1 year ago

Verified 8d94219a

convert : fix broken sentencepiece vocab (#14416)

CISC committed 1 year ago

Verified f667f1e6

Merge branch 'master' into compilade/test-model-random

compilade committed 1 year ago

7c3f9c22

model : gemma3n text-only (#14400)

ngxson committed 1 year ago

Verified 8846aace

cmake: regen vulkan shaders when shaders-gen sources change (#14398)

mtmcp committed 1 year ago

Verified a01047b0

llama : return mistral-v7-tekken as default template only (#14390)

CISC committed 1 year ago

Verified b2534622

metal : add special-case mat-vec mul for ne00 == 4 (#14385)

ggerganov committed 1 year ago

Verified e8215dbb

metal : batch rows copy in a single threadgroup (#14384)

ggerganov committed 1 year ago

Verified 5783ae43

docs: update s390x documentation + add faq (#14389)

taronaeo committed 1 year ago

Verified bf5bcd0b

musa: enable fp16 mma (all) and cublas on qy2 (#13842)

yeahdongcn committed 1 year ago

Verified 716301d1

ggml-cpu: enable IBM NNPA Vector Intrinsics (#14317)

taronaeo committed 1 year ago

Verified 60ef23d6

ggml : do not output unprintable characters on GGUF load failure (#14381)

CISC committed 1 year ago

Verified b193d530

sycl: GGML_SYCL_DISABLE_OPT on by default for all Intel Devices (#13973)

ShanoToni committed 1 year ago

Verified 2bf9d539

opencl: ref count `ggml_backend_opencl_context` and refactor profiling (#14254)

lhez committed 1 year ago

Verified 73e53dc8

batch : fix check for empty sequences in memory (#14364)

ggerganov committed 1 year ago

Verified 62af4642

cmake : use LLAMA_BUILD_NUMBER when defining LLAMA_INSTALL_VERSION (#14362)

mbaudier committed 1 year ago

Verified c148cf19

server : move no API key doc to /health (#14352)

pnb committed 1 year ago

Verified 1b809cee

main : honor --verbose-prompt on interactive prompts (#14350)

CISC committed 1 year ago

Verified abf24104

jinja : Add Mistral-Small-3.2-24B-Instruct-2506.jinja (#14349)

bartowski1182 committed 1 year ago

Verified 901e20bb

CUDA/HIP: optimize mmv paths taken for HIP devices (#14324)

IMbackK committed 1 year ago

Verified 0142961a

ci: add workflow for relocatable cmake package (#14346)

mtmcp committed 1 year ago

Verified ce82bd01

vulkan: update windows SDK in release.yml (#14344)

jeffbolznv committed 1 year ago

Verified bf2a99e3

llama : better rwkv chat template and add missing `inputs.use_jinja` setting (#14336)

MollySophia committed 1 year ago

Verified 72c6bc3f

CUDA: mul_mat_v support for batch sizes > 1 (#14262)

JohannesGaessler committed 1 year ago

Verified defe2158

kv-cells : fix tracking of seq_pos (#14339)

ggerganov committed 1 year ago

Verified 7b50d589

vulkan: update windows SDK in CI (#14334)

jeffbolznv committed 1 year ago

Verified 3a9457df

quantize : handle user-defined pruning of whole layers (blocks) (#13037)

EAddario committed 1 year ago

Verified fa4a9f2a

gguf-py : fix SpecialVocab parsing when post_processor is null (#14330)

CISC committed 1 year ago

Verified 238005c2

Newer Older