ngxson/llama.cpp

Pull Requests Commits

server : fix cache_tokens bug with no cache_prompt

ngxson committed 239 days ago

8f3c6fab

webui : use fflate for more deterministic gzip compress (#13525)

ngxson committed 239 days ago

Verified bb1681fb

webui: Allow pasting file from clipboard (#13526)

luca020400 committed 239 days ago

Verified d486dd3e

docs: Update link to ggml-org in multimodal.md (#13513)

ddpasa committed 239 days ago

Verified 21ca987f

scripts : fix compare-llama-bench.py show parameter (#13514)

CISC committed 239 days ago

Verified be1d4a13

vulkan: workaround FA compile failures on macos (#13517)

jeffbolznv committed 239 days ago

Verified ab3971f2

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

EAddario committed 240 days ago

Verified e5c834f7

clip : clip.h become private API (⚠️ breaking change) (#13510)

ngxson committed 240 days ago

Verified 71bdbdb5

metal : use FA-vec kernel up to batch size 20 (#13496)

ggerganov committed 240 days ago

Verified f0995d28

metal : optimize multi-sequence FA vec kernel (#13493)

ggerganov committed 240 days ago

Verified c252e0c4

ggml-cpu: Update KleidiAI to v1.6 and fix include directives (#13509)

eddnjjn committed 240 days ago

Verified 4f711afe

batched-bench : fix pp batch contents (#13492)

ggerganov committed 240 days ago

Verified b89d605a

mtmd : remove libllava, remove clip-quantize-cli (⚠️ breaking change) (#13460)

ngxson committed 240 days ago

Verified b4726345

scripts : support arbitrary input file formats in compare-llama-bench.py (#13455)

CISC committed 240 days ago

Verified bf793711

model : Granite MoE shared (#13269)

gabe-l-hart committed 240 days ago

Verified d590cd4c

ggerganov committed 240 days ago

1e2809bc

llama-bench : add defrag-thold, check for invalid ranges (#13487)

slaren committed 240 days ago

Verified cf0a43bb

opencl: remove unnecessary assert for `add` (#13257)

lhez committed 241 days ago

Verified f0d46ef1

clip : cap max image size 1024 for qwen vl model (#13478)

ngxson committed 241 days ago

Verified de4c07f9

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

JohannesGaessler committed 241 days ago

Verified 10d2af0e

context : fix state io for memory-less contexts (#13470)

ggerganov committed 241 days ago

Verified 064cc596

server : allow content to be null in oaicompat_completion_params_parse (#13477)

anudit committed 241 days ago

Verified 91159ee9

llama-bench : accept ranges for integer parameters (#13410)

slaren committed 241 days ago

Verified 22cdab34

ggml-cpu: Integrate fp32=bf16xbf16 SME KleidiAI kernel (#13053)

eddnjjn committed 241 days ago

Verified a71a4075

CUDA: fix misaligned synchronization in FA (#13469)

JohannesGaessler committed 241 days ago

Verified 95e18884

ggml : add mrope kernel for metal (#13457)

ngxson committed 241 days ago

Verified df849192

enable dpcpp nightly builds with libraries (#13406)

AD2605 committed 241 days ago

Verified 14492144

mtmd : Use RMS norm for InternVL 3 38B and 78B mmproj (#13459)

city96 committed 241 days ago

Verified c1040239

tools : fix uninitialized llama_batch in server (#13436)

aumfer committed 242 days ago

Verified 9a390c48

scripts : exit compare-llama-bench.py gracefully when there's nothing to compare (#13451)

CISC committed 242 days ago

Verified 09232370

Older