ggerganov/llama.cpp

Pull Requests Commits
spec : remove check rate (#19377)
srogmann committed 112 days ago
Verified 292f6908
ci : add metal server workflows (#19293)
ggerganov committed 112 days ago
Verified 81ddc60c
revert : "[Model] Qwen3.5 dense and MoE support (no vision) (#19435)" (#19453)
ggerganov committed 112 days ago
Verified 972f323e
ggml-virtgpu: add backend documentation (#19354)
kpouget committed 112 days ago
Verified f5e7734f
cmake : add variable to skip installing tests (#19370)
WhyNotHugo committed 112 days ago
Verified 1e8924fd
[Model] Qwen3.5 dense and MoE support (no vision) (#19435)
pwilkin committed 113 days ago
Verified 39bf692a
CUDA: Fix non-contig rope (#19338)
ORippler committed 113 days ago
Verified e06088da
rpc : update from common.cpp (#19400)
angt committed 113 days ago
Verified 5fa1c190
server : improve context checkpoint logic (#19408)
ggerganov committed 113 days ago
Verified eb449cdf
llama-quantize : cleanup `--help` output (#19317)
ddh0 committed 113 days ago
Verified 5999b50e
ci : remove server job from webui and move slow test (#19424)
CISC committed 114 days ago
Verified 9a5f5779
ci : use -j param correctly when building with sanitizers (#19411)
ggerganov committed 114 days ago
Verified 96441c95
metal : consolidate bin kernels (#19390)
ggerganov committed 114 days ago
Verified 8872ad21
metal : fix event synchronization in cpy_tensor_async (#19402)
ggerganov committed 114 days ago
Verified 34ba7b5a
model : support Step3.5-Flash (#19283)
forforever73 committed 115 days ago
Verified b8311181
gguf-py : bump sentencepiece version (#19319)
Ahajha committed 115 days ago
Verified 3228e772
ggml-webgpu: JIT compile binary operators and handle binding overlaps (#19310)
abhijitramesh committed 115 days ago
Verified 7fbd36c5
sycl: add F16 support for GGML_OP_CEIL (#19306)
NechamaKrashinski committed 115 days ago
Verified 537eadb1
tests: reduce number of FA test permutations (#19381)
jeffbolznv committed 115 days ago
Verified db6adb3c
common : add common_speculative_is_compat() (#19270)
ggerganov committed 115 days ago
Verified dfde5993
unicode : MSVC regex fix (#19340)
Iemand005 committed 115 days ago
Verified 06bf3796
Kimi-Linear support (backend agnostic + MLA KV cache) (#18755)
ymcki committed 115 days ago
Verified 3688c4f5
vulkan: For coopmat2 FA, use fp16 accumulators for the final result (#19376)
jeffbolznv committed 115 days ago
Verified 1946e46f
vulkan: make FA mask/softcap enables spec constants (#19309)
jeffbolznv committed 115 days ago
Verified f9bd518a
metal : skip loading all-zero mask (#19337)
ggerganov committed 115 days ago
Verified 7fcf1ef4
llama : rename llama-sampling to llama-sampler (#19363)
danbev committed 115 days ago
Verified e696cfc0
cuda : cuda graphs now compare all node params (#19383)
ggerganov committed 115 days ago
Verified 3e216476
metal : adaptive CPU/GPU interleave based on number of nodes (#19369)
ggerganov committed 116 days ago
Verified 22cae832
vulkan: Preprocess FA mask to detect all-neg-inf and all-zero. (#19281)
jeffbolznv committed 116 days ago
Verified 449ec2ab
benches : update models + numbers (#19359)
ggerganov committed 116 days ago
Verified 3795cc1e
Terms Of Use Privacy Policy Refund Policy Impressum