ggerganov/llama.cpp

Pull Requests Commits
vulkan: sparse flash attention for quantized K/V (#29639)
fxgsell committed 20 days ago
Verified b3daa077
llama : fix unexpected graph reallocation in the k-pool models (#29958)
ggerganov committed 20 days ago
Verified c173a53b
kv-cache: fix restoring mismatched KV cache rotation by saving exact rotation metadata (#28498)
eapache committed 20 days ago
Verified 21079106
ci : winget urls must be separate strings (#29978)
CISC committed 20 days ago
Verified e5983d67
ci : fix docker workflow permissions (#29979)
CISC committed 20 days ago
Verified 4ca6b76f
ggml-cpu : add Q8_0 IME1 matrix kernel for SpacemiT X60 (#28479)
alanhc committed 20 days ago
Verified 9f12cd4a
vulkan : Fix undeclared identifiers when -DGGML_VULKAN_RUN_TESTS=ON (#29912)
EAddario committed 20 days ago
Verified ebe18bee
webgpu: add MMVQ support for Q1_0/Q5_0/Q5_1/Q3_K/Q5_K/Q6_K/MXFP4 (#29483)
yomaytk committed 20 days ago
Verified 8216c846
cuda: tile the lightning indexer over keys and tokens for 4 heads (#29901)
ServeurpersoCom committed 20 days ago
Verified 1b43d311
metal : few-row MMA mat-mul (#29869)
pratiknarola-t committed 20 days ago
Verified a3a1c474
CUDA: use MMVF for thin f16/bf16 mul_mat at small batch size (#29633)
ynankani committed 20 days ago
Verified 9d3aba6b
CUDA: prefer whole-tile FlashAttention scheduling for efficient two-stage kernels (#29435)
anujj committed 20 days ago
Verified d89651a7
log, server: self contained colors, split child commands from logs in router mode (#29895)
ServeurpersoCom committed 20 days ago
Verified a7fb71fa
llama: support both embd + raw tokens in batch (#29622)
ngxson committed 20 days ago
Verified 0bb496db
CUDA: refactor swizzling code (#29612)
JohannesGaessler committed 20 days ago
Verified 2ca15f54
ggml-cpu: support BF16/FP16/FP32 K tails in tinyBLAS on x86 (#29806)
SongXiaoXi committed 20 days ago
Verified a7b94df2
cuda : move neu_padded to where it is used (#29940)
angt committed 20 days ago
Verified 0eb6d9a8
ci : windows llvm build requires ninja multi-config (#29959)
CISC committed 20 days ago
Verified 2e7c58c5
ci : add windows arm64 vulkan release (#29954)
CISC committed 20 days ago
Verified 7f2dd88b
AGENTS.md : revamp (#29656)
am17an committed 20 days ago
Verified bf79dbbc
chat-peg-parser : clear current_tool when pending_tool_call is reset (#29942)
T-Anas committed 20 days ago
Verified dbe4c3ed
ci : set default permissions (#29945)
CISC committed 20 days ago
Verified 46847e61
cuda : move blocks_per_col to where it is used (#29939)
angt committed 20 days ago
Verified 2bc56357
CUDA: fix MMQ memory fault if n_expert >> n_ubatch (#29941)
JohannesGaessler committed 21 days ago
Verified dd266785
vulkan: fix rdna4 mat_vec tuning (#29934)
0cc4m committed 21 days ago
Verified 16c163d5
imatrix: calculate activation-based statistics for new format (GGUF) imatrices (#14891)
EAddario committed 21 days ago
Verified 05043961
spec : fix n-gram drafts rejected at temp > 0 after truncation (#29924)
praneshgo committed 21 days ago
Verified 8330e969
common : prepare load_from_models_dir() for path conversion (#29674)
angt committed 21 days ago
Verified 6716df69
server : fix dead LLAMA_ARG_HF_REPO_FILE key in preset allow-list (#29938)
angt committed 21 days ago
Verified bf9a0ccc
ci : pushing tag needs deploy key (#29937)
CISC committed 21 days ago
Verified 0faee500
Terms Of Use Privacy Policy Refund Policy Impressum