ggerganov/llama.cpp

Pull Requests Commits
vulkan: optimizations for direct convolution (#14933)
jeffbolznv committed 160 days ago
Verified a9f7541e
CUDA: fix MMQ nwarps for AMD with warp_size==32 (#15014)
JohannesGaessler committed 161 days ago
Verified 9c35706b
vendor : update vendored copy of google/minja (#15011)
l-austenfeld committed 161 days ago
Verified c76b420e
model : add hunyuan dense (#14878)
stevenkuang-tencent committed 161 days ago
Verified 0f5ccd6f
opencl: add f16 for `add`, `sub`, `mul`, `div` (#14984)
lhez committed 161 days ago
Verified 1c872f71
ggml : Q2k interleaving implementation - x86/x64 SIMD (#14373)
Srihari-mcw committed 162 days ago
Verified baad9488
graph : fix equal_seq() check (#14986)
ggerganov committed 162 days ago
Verified ba42794c
docker : add cann build pipline (#14591)
diannaojiang committed 162 days ago
Verified 2860d479
compare-commits.sh: support both llama-bench and test-backend-ops (#14392)
yeahdongcn committed 162 days ago
Verified 484b2091
quantize : skip tensor override when in fallback mode (#14995)
EAddario committed 162 days ago
Verified daf2dd78
llama : add simple option to enable CPU for MoE weights (--cpu-moe) (#14992)
slaren committed 162 days ago
Verified a06ed5fe
Fix params bug in diffusion example (#14993)
am17an committed 162 days ago
Verified 78452405
llama : allow other bufts when overriding to CPU, add --no-repack option (#14990)
slaren committed 162 days ago
Verified d6818d06
Vulkan: Fix minor debug mode issues (#14899)
0cc4m committed 162 days ago
Verified e08a9882
mtmd : support MiniCPM-V 4.0 (#14983)
tc-mb committed 162 days ago
Verified 952a47f4
MODEL_TENSOR.SSM_DT_NORM has defined twice (#14991)
csabakecskemeti committed 162 days ago
Verified 36e5fe7b
server : implement universal assisted decoding (#12635)
g2mt committed 162 days ago
Verified 94933c8c
llama : merge build_moe_ffn_from_probs function into build_moe_ffn (#14968)
wdl339 committed 162 days ago
Verified c1dacaa9
server : add openai-style logit_bias support (#14946)
lukasstraub2 committed 162 days ago
Verified a9f77a8b
Add LLaDA 8b Diffusion model (#14771)
am17an committed 162 days ago
Verified 8a4a8562
CANN: Improve loading efficiency after converting weights to NZ format. (#14985)
hipudding committed 162 days ago
Verified 11490b36
graph : reduce splits for recurrent and hybrid models (#14825)
compilade committed 163 days ago
Verified 66625a59
opencl: add `mul_mat_f32_f32_l4_lm` and `mul_mat_f16_f32_l4_lm` (#14809)
lhez committed 163 days ago
Verified 6e672545
quantize : fix using combined imatrix GGUFs (multiple datasets) (#14973)
EAddario committed 163 days ago
Verified e9192bec
server : add support for `embd_normalize` parameter (#14964)
danbev committed 163 days ago
Verified 41e78c56
HIP: enable mfma mmq on gfx908 and gfx90a for select datatypes and shapes (#14949)
IMbackK committed 163 days ago
Verified ad4a7001
sync : ggml
ggerganov committed 163 days ago
e32a4ec6
cmake : Fix BLAS link interface (ggml/1316)
dg0yt committed 163 days ago
e228de94
vulkan : fix 32-bit builds (ggml/1313)
dg0yt committed 163 days ago
73a8e5ca
CUDA: skip masked KV slices for all FA kernels (#14924)
JohannesGaessler committed 163 days ago
Verified 92b8810e
Terms Of Use Privacy Policy Refund Policy Impressum