ggerganov/llama.cpp

Pull Requests Commits
llama/ggml: add LLM training support (#10544)
JohannesGaessler committed 300 days ago
Verified 10d2af0e
context : fix state io for memory-less contexts (#13470)
ggerganov committed 300 days ago
Verified 064cc596
server : allow content to be null in oaicompat_completion_params_parse (#13477)
anudit committed 300 days ago
Verified 91159ee9
llama-bench : accept ranges for integer parameters (#13410)
slaren committed 300 days ago
Verified 22cdab34
ggml-cpu: Integrate fp32=bf16xbf16 SME KleidiAI kernel (#13053)
eddnjjn committed 300 days ago
Verified a71a4075
CUDA: fix misaligned synchronization in FA (#13469)
JohannesGaessler committed 300 days ago
Verified 95e18884
ggml : add mrope kernel for metal (#13457)
ngxson committed 300 days ago
Verified df849192
enable dpcpp nightly builds with libraries (#13406)
AD2605 committed 301 days ago
Verified 14492144
mtmd : Use RMS norm for InternVL 3 38B and 78B mmproj (#13459)
city96 committed 301 days ago
Verified c1040239
tools : fix uninitialized llama_batch in server (#13436)
aumfer committed 301 days ago
Verified 9a390c48
scripts : exit compare-llama-bench.py gracefully when there's nothing to compare (#13451)
CISC committed 301 days ago
Verified 09232370
CUDA: fix crash with partial offloading of MoE (#13439)
JohannesGaessler committed 301 days ago
Verified 7474e00b
Add `--no-op-offload` to improve `-ot` pp perf in MoE models like llama4 400B (#13386)
hjc4869 committed 301 days ago
Verified 7f323a58
mtmd : support InternVL 3 38B and 78B mmproj (#13443)
city96 committed 301 days ago
Verified 3eac2093
mtmd : move helpers to dedicated file (#13442)
ngxson committed 301 days ago
Verified a634d75d
docs : Fix typo in InternVL3 model name (#13440)
99991 committed 302 days ago
Verified 62d4250e
CUDA: fix race conditions FlashAttention kernels (#13438)
JohannesGaessler committed 302 days ago
Verified 0208355f
vocab : add ByteDance-Seed/Seed-Coder (#13423)
CISC committed 302 days ago
Verified d2a4ef05
mtmd : add hard limit on image resolution for qwen2vl / qwen2.5vl (#13434)
ngxson committed 302 days ago
Verified 15e6125a
server : update docs (#13432)
ngxson committed 302 days ago
Verified 3b24d26c
llguidance : set tokenizer slices to default (#13424)
CISC committed 302 days ago
Verified 43dfd741
ci: free_disk_space flag enabled for intel variant (#13426)
Thammachart committed 302 days ago
Verified b064a51a
mtmd : support InternVL 2.5 and 3 (#13422)
ngxson committed 302 days ago
Verified 053367d1
CUDA: fix FlashAttention on Turing (#13415)
JohannesGaessler committed 303 days ago
Verified d8919424
arg : add env var to control mmproj (#13416)
ngxson committed 303 days ago
Verified 7fef1176
vulkan: scalar flash attention implementation (#13324)
jeffbolznv committed 303 days ago
Verified dc1d2adf
chore(llguidance): use tagged version that does not break the build (#13413)
HRKings committed 303 days ago
Verified 7c28a74e
server : vision support via libmtmd (#12898)
ngxson committed 303 days ago
Verified 33eff402
sycl : implementation of reordered Q4_0 MMVQ for Intel GPUs (#12858)
Alberto Cabrera Pérez committed 303 days ago
Verified 17512a94
metal : optimize MoE for large batches (#13388)
ggerganov committed 303 days ago
Verified 611aa914
Terms Of Use Privacy Policy Refund Policy Impressum