ggml-org/llama.cpp

Pull Requests Commits
MODEL_TENSOR.SSM_DT_NORM has defined twice (#14991)
csabakecskemeti committed 297 days ago
Verified 36e5fe7b
server : implement universal assisted decoding (#12635)
g2mt committed 297 days ago
Verified 94933c8c
llama : merge build_moe_ffn_from_probs function into build_moe_ffn (#14968)
wdl339 committed 297 days ago
Verified c1dacaa9
server : add openai-style logit_bias support (#14946)
lukasstraub2 committed 297 days ago
Verified a9f77a8b
Add LLaDA 8b Diffusion model (#14771)
am17an committed 297 days ago
Verified 8a4a8562
CANN: Improve loading efficiency after converting weights to NZ format. (#14985)
hipudding committed 297 days ago
Verified 11490b36
graph : reduce splits for recurrent and hybrid models (#14825)
compilade committed 297 days ago
Verified 66625a59
opencl: add `mul_mat_f32_f32_l4_lm` and `mul_mat_f16_f32_l4_lm` (#14809)
lhez committed 297 days ago
Verified 6e672545
quantize : fix using combined imatrix GGUFs (multiple datasets) (#14973)
EAddario committed 298 days ago
Verified e9192bec
server : add support for `embd_normalize` parameter (#14964)
danbev committed 298 days ago
Verified 41e78c56
HIP: enable mfma mmq on gfx908 and gfx90a for select datatypes and shapes (#14949)
IMbackK committed 298 days ago
Verified ad4a7001
sync : ggml
ggerganov committed 298 days ago
e32a4ec6
cmake : Fix BLAS link interface (ggml/1316)
dg0yt committed 298 days ago
e228de94
vulkan : fix 32-bit builds (ggml/1313)
dg0yt committed 298 days ago
73a8e5ca
CUDA: skip masked KV slices for all FA kernels (#14924)
JohannesGaessler committed 298 days ago
Verified 92b8810e
tests : update for LLAMA_SET_ROWS=1 (#14961)
ggerganov committed 298 days ago
Verified 00131d6e
graph : fix stack-use-after-return (#14960)
ggerganov committed 298 days ago
Verified 1e15bfd4
embeddings: fix extraction of CLS pooling results (#14927)
iamlemec committed 298 days ago
Verified a118d802
CANN: update ops docs (#14935)
bachelor-dou committed 298 days ago
Verified 61550f82
HIP: remove the use of __HIP_PLATFORM_AMD__, explicitly support only AMD targets (#14945)
IMbackK committed 299 days ago
Verified aa79524c
HIP: add GGML_HIP_MMQ_MFMA option to allow disableing the MFMA path. (#14930)
IMbackK committed 299 days ago
Verified b77d1117
HIP: Ignore unsupported unroll transformation in fattn-vec (#14931)
IMbackK committed 299 days ago
Verified c7aa1364
common : avoid logging partial messages (which can contain broken UTF-8 sequences) (#14937)
kallewoof committed 299 days ago
Verified 1a67fcc3
CANN: Add ggml_set_rows (#14943)
hipudding committed 299 days ago
Verified 204f2cf1
cuda : add softcap fusion (#14907)
CISC committed 299 days ago
Verified 138b288b
server-bench: make seed choice configurable (#14929)
JohannesGaessler committed 299 days ago
Verified bbd0f917
CUDA: add roll (#14919)
am17an committed 299 days ago
Verified 0a5036be
opencl : add ops docs (#14910)
lhez committed 300 days ago
Verified 8ad7b3e6
test-backend-ops : extend test case filtering (#14865)
tlemo committed 300 days ago
Verified bda62193
llama-bench : use local GPUs along with RPC servers (#14917)
rgerganov committed 300 days ago
Verified c556418b
Terms Of Use Privacy Policy Refund Policy Impressum