ggerganov/llama.cpp

Pull Requests Commits
CUDA: fix replacment of bad archs in CMake (#18457)
JohannesGaessler committed 23 days ago
Verified 0bd1212a
server : Cmdline arg -to changes http read timeout from current 600sec default (#18279)
wbtek committed 23 days ago
Verified 5b1248c9
contributing: tighten AI usage policy (#18388)
ngxson committed 23 days ago
Verified 3595ae59
android: routine maintenance - Dec 2025 (#18338)
naco-siren committed 23 days ago
Verified c1366056
server : handle closed connection for tasks (#18459)
ggerganov committed 23 days ago
Verified 2a85f720
model-conversion : add device option to embd run orig model (#18386)
danbev committed 23 days ago
Verified 7cbec34a
retrieval : use at most n_seq_max chunks (#18400)
hectorem2 committed 23 days ago
Verified 0c898640
common: fix return value check for setpriority (#18412)
o7si committed 23 days ago
Verified daa242df
CUDA: Blackwell features for non-native builds (#18436)
JohannesGaessler committed 23 days ago
Verified e70e640d
cuda: fix race condition in cumsum (#18448)
am17an committed 23 days ago
Verified 5fa66c6e
ci : re-enable rocm build on amd64 (#18439)
TimNN committed 23 days ago
Verified 382808c1
HIP: Use mmq on MFMA devices for MUL_MAT_ID in cases where a lot of splits would be generated (#18202)
IMbackK committed 24 days ago
Verified 4ffc47cb
model : Plamo3 support (#17304)
mmnga committed 24 days ago
Verified 9c675c71
Revert "ggml-cuda: use CMAKE_CUDA_ARCHITECTURES if set when GGML_NATIVE=ON (#18413)" (#18426)
am17an committed 24 days ago
Verified 07a0c4ba
rpc: fix segfault on invalid endpoint format (#18387)
o7si committed 24 days ago
Verified 60f17f56
llama-fit-params: fix step size for last device (#18415)
JohannesGaessler committed 24 days ago
Verified f8d561eb
github: update issue templates [no ci] (#18410)
JohannesGaessler committed 24 days ago
Verified e59efe6a
mtmd: clarify that we no longer accept AI-generated PRs (#18406)
ngxson committed 24 days ago
Verified cffa5c46
cmake: Added more x86_64 CPU backends when building with `GGML_CPU_ALL_VARIANTS=On` (#18186)
bberberov committed 24 days ago
Verified 94de74e7
ggml-cuda: use CMAKE_CUDA_ARCHITECTURES if set when GGML_NATIVE=ON (#18413)
QDelta committed 24 days ago
Verified 4fd59e84
opencl: allow resizing transpose buffers (#18384)
lhez committed 24 days ago
Verified 08566977
llama-fit-params: fix overflow check (#18354)
JohannesGaessler committed 25 days ago
Verified a4bf3588
llama: fix magic number of 999 for GPU layers (#18266)
JohannesGaessler committed 25 days ago
Verified 026d2ad4
ggml-cuda: Use same regex for GGML_NATIVE=OFF (#18407)
am17an committed 25 days ago
Verified 06705fdc
llama_fit_params: return enum for fail vs. error (#18374)
JohannesGaessler committed 25 days ago
Verified a52dc60b
llama-fit-params: fix Gemma 3 calculation (#18372)
JohannesGaessler committed 25 days ago
Verified 9045c9af
vulkan: preprocess mul_mat_id experts and discard workgroups more quickly (#18352)
jeffbolznv committed 25 days ago
Verified c9ced491
vulkan: optimize decodeFuncB in coopmat2 mul_mat_id shader (#18349)
jeffbolznv committed 26 days ago
Verified 7ac89021
vulkan: Use BK=32 for coopmat2 mul_mat_id (#18332)
jeffbolznv committed 26 days ago
Verified 9bf20d8a
vulkan: small dequantization improvements (#18380)
netrunnereve committed 26 days ago
Verified cb999704
Terms Of Use Privacy Policy Refund Policy Impressum