Skip to content

Actions: ngxson/llama.cpp

Actions

Release

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
release.yml
965 workflow runs
965 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

metal : use FA-vec kernel up to batch size 20 (#13496) Release #40: Commit f0995d2 pushed by ngxson
41m 11s master
master
41m 11s
mtmd : remove libllava, remove clip-quantize-cli (⚠️ breaking change)… Release #39: Commit b472634 pushed by ngxson
32m 12s master
master
32m 12s
model : Granite MoE shared (#13269) Release #38: Commit d590cd4 pushed by ngxson
32m 39s master
master
32m 39s
llama-bench : add defrag-thold, check for invalid ranges (#13487) Release #37: Commit cf0a43b pushed by ngxson
21m 7s master
master
21m 7s
opencl: remove unnecessary assert for add (#13257) Release #36: Commit f0d46ef pushed by ngxson
32m 13s master
master
32m 13s
clip : cap max image size 1024 for qwen vl model (#13478) Release #35: Commit de4c07f pushed by ngxson
1h 4m 34s master
master
1h 4m 34s
llama/ggml: add LLM training support (#10544) Release #34: Commit 10d2af0 pushed by ngxson
54m 6s master
master
54m 6s
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
ProTip! You can narrow down the results and go further in time using created:<2025-05-10 or the other filters available.