Skip to content

Actions: ngxson/llama.cpp

Actions

Release

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
release.yml
1,058 workflow runs
1,058 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

context : fix state io for memory-less contexts (#13470) Release #33: Commit 064cc59 pushed by ngxson
42m 23s master
master
42m 23s
server : allow content to be null in oaicompat_completion_params_pars… Release #32: Commit 91159ee pushed by ngxson
59m 56s master
master
59m 56s
llama-bench : accept ranges for integer parameters (#13410) Release #31: Commit 22cdab3 pushed by ngxson
56m 26s master
master
56m 26s
CUDA: fix misaligned synchronization in FA (#13469) Release #30: Commit 95e1888 pushed by ngxson
22m 1s master
master
22m 1s
ggml : add mrope kernel for metal (#13457) Release #29: Commit df84919 pushed by ngxson
17m 18s master
master
17m 18s
enable dpcpp nightly builds with libraries (#13406) Release #28: Commit 1449214 pushed by ngxson
27m 2s master
master
27m 2s
mtmd : Use RMS norm for InternVL 3 38B and 78B mmproj (#13459) Release #27: Commit c104023 pushed by ngxson
19m 30s master
master
19m 30s
tools : fix uninitialized llama_batch in server (#13436) Release #26: Commit 9a390c4 pushed by ngxson
25m 48s master
master
25m 48s
CUDA: fix crash with partial offloading of MoE (#13439) Release #25: Commit 7474e00 pushed by ngxson
28m 42s master
master
28m 42s
Add --no-op-offload to improve -ot pp perf in MoE models like lla… Release #24: Commit 7f323a5 pushed by ngxson
52m 49s master
master
52m 49s
mtmd : support InternVL 3 38B and 78B mmproj (#13443) Release #23: Commit 3eac209 pushed by ngxson
30m 12s master
master
30m 12s
CUDA: fix race conditions FlashAttention kernels (#13438) Release #22: Commit 0208355 pushed by ngxson
27m 24s master
master
27m 24s
vocab : add ByteDance-Seed/Seed-Coder (#13423) Release #21: Commit d2a4ef0 pushed by ngxson
34m 37s master
master
34m 37s
mtmd : add hard limit on image resolution for qwen2vl / qwen2.5vl (#1… Release #20: Commit 15e6125 pushed by ngxson
22m 1s master
master
22m 1s
llguidance : set tokenizer slices to default (#13424) Release #19: Commit 43dfd74 pushed by ngxson
32m 57s master
master
32m 57s
mtmd : support InternVL 2.5 and 3 (#13422) Release #18: Commit 053367d pushed by ngxson
20m 18s master
master
20m 18s
CUDA: fix FlashAttention on Turing (#13415) Release #17: Commit d891942 pushed by ngxson
25m 2s master
master
25m 2s
arg : add env var to control mmproj (#13416) Release #16: Commit 7fef117 pushed by ngxson
27m 12s master
master
27m 12s
vulkan: scalar flash attention implementation (#13324) Release #15: Commit dc1d2ad pushed by ngxson
25m 54s master
master
25m 54s
chore(llguidance): use tagged version that does not break the build (… Release #14: Commit 7c28a74 pushed by ngxson
24m 8s master
master
24m 8s
server : vision support via libmtmd (#12898) Release #13: Commit 33eff40 pushed by ngxson
29m 2s master
master
29m 2s
sycl : implementation of reordered Q4_0 MMVQ for Intel GPUs (#12858) Release #12: Commit 17512a9 pushed by ngxson
30m 56s master
master
30m 56s
metal : optimize MoE for large batches (#13388) Release #11: Commit 611aa91 pushed by ngxson
38m 25s master
master
38m 25s
CUDA: FA support for Deepseek (Ampere or newer) (#13306) Release #10: Commit 0cf6725 pushed by ngxson
56m 46s master
master
56m 46s
llama : do not crash if there is no CPU backend (#13395) Release #9: Commit 27ebfca pushed by ngxson
27m 29s master
master
27m 29s
ProTip! You can narrow down the results and go further in time using created:<2025-05-09 or the other filters available.