ggml-org/llama.cpp

Pull Requests Commits
sycl: reordered Q4_K MMVQ (#13109)
sgeor255 committed 224 days ago
Verified 64bb51cf
sycl: use oneDNN for matrices multiplication (#12972)
lslusarczyk committed 224 days ago
Verified 9c404ed5
llama-bench : fix -ot with dl backends (#13563)
slaren committed 224 days ago
Verified 6c8b9150
webui : handle PDF input (as text or image) + convert pasted long content to file (#13562)
ngxson committed 224 days ago
Verified 3cc1f1f1
server : proper error handling for missing elements in messages array (OpenAI compatible backend) (#13540)
pwilkin committed 224 days ago
Verified c753d7be
bench : handle decode errors (#13548)
ggerganov committed 224 days ago
Verified b2838049
`server`: inject date_string in llama 3.x template + fix date for firefunction v2 (#12802)
ochafik committed 224 days ago
Verified aa48e373
kv-cache : fix out-of-bounds view during reserve graph (#13547)
ggerganov committed 224 days ago
Verified e3a9421b
arm64: optimize q6_k_q8_k kernel with i8mm (#13519)
cyb70289 committed 225 days ago
Verified 5ab5d5fb
`common`: add partial regex support (#12808)
ochafik committed 225 days ago
Verified 3198405e
editorconfig : fix trailing whitespace from #13542 (#13546)
CISC committed 225 days ago
Verified f5170c1d
fix: crash when calling `llama_state_get_size` on a context without a KV cache (#13542)
giladgd committed 225 days ago
Verified 017f10b5
CUDA: fix crash on large batch size for quant. MoE (#13537)
JohannesGaessler committed 225 days ago
Verified 4696d567
llama : fix quantize with dl backends (#13539)
slaren committed 225 days ago
Verified b7d26720
CUDA: faster Deepseek FA, add Turing support (#13435)
JohannesGaessler committed 225 days ago
Verified 6da34fa2
fix: Move build_inp_pos to the top of the graph section for build_granite (#13538)
gabe-l-hart committed 225 days ago
Verified 5e7d95e2
server : passthrough the /models endpoint during loading (#13535)
ggerganov committed 225 days ago
Verified 05317443
server : fix cache_tokens bug with no cache_prompt (#13533)
ngxson committed 225 days ago
Verified 360a9c98
cmake: simplify vulkan shader test logic (#13263)
bandoti committed 225 days ago
Verified 09d13d94
vulkan: KHR_coopmat flash attention (#13506)
jeffbolznv committed 225 days ago
Verified 24e86cae
webui : use fflate for more deterministic gzip compress (#13525)
ngxson committed 225 days ago
Verified bb1681fb
webui: Allow pasting file from clipboard (#13526)
luca020400 committed 225 days ago
Verified d486dd3e
docs: Update link to ggml-org in multimodal.md (#13513)
ddpasa committed 225 days ago
Verified 21ca987f
scripts : fix compare-llama-bench.py show parameter (#13514)
CISC committed 225 days ago
Verified be1d4a13
vulkan: workaround FA compile failures on macos (#13517)
jeffbolznv committed 225 days ago
Verified ab3971f2
quantize : improve tensor-type pattern matching (#13033)
EAddario committed 226 days ago
Verified e5c834f7
clip : clip.h become private API (⚠️ breaking change) (#13510)
ngxson committed 226 days ago
Verified 71bdbdb5
metal : use FA-vec kernel up to batch size 20 (#13496)
ggerganov committed 226 days ago
Verified f0995d28
metal : optimize multi-sequence FA vec kernel (#13493)
ggerganov committed 226 days ago
Verified c252e0c4
ggml-cpu: Update KleidiAI to v1.6 and fix include directives (#13509)
eddnjjn committed 226 days ago
Verified 4f711afe
Terms Of Use Privacy Policy Refund Policy Impressum