Skip to content

Actions: ngxson/llama.cpp

Actions

Release

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
release.yml
1,152 workflow runs
1,152 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

fix: crash when calling llama_state_get_size on a context without a… Release #52: Commit 017f10b pushed by ngxson
27m 45s master
master
27m 45s
CUDA: fix crash on large batch size for quant. MoE (#13537) Release #51: Commit 4696d56 pushed by ngxson
35m 4s master
master
35m 4s
llama : fix quantize with dl backends (#13539) Release #50: Commit b7d2672 pushed by ngxson
49m 32s master
master
49m 32s
CUDA: faster Deepseek FA, add Turing support (#13435) Release #49: Commit 6da34fa pushed by ngxson
41m 34s master
master
41m 34s
fix: Move build_inp_pos to the top of the graph section for build_gra… Release #48: Commit 5e7d95e pushed by ngxson
40m 13s master
master
40m 13s
server : passthrough the /models endpoint during loading (#13535) Release #47: Commit 0531744 pushed by ngxson
38m 37s master
master
38m 37s
server : fix cache_tokens bug with no cache_prompt (#13533) Release #46: Commit 360a9c9 pushed by ngxson
25m 41s master
master
25m 41s
cmake: simplify vulkan shader test logic (#13263) Release #45: Commit 09d13d9 pushed by ngxson
33m 11s master
master
33m 11s
vulkan: KHR_coopmat flash attention (#13506) Release #44: Commit 24e86ca pushed by ngxson
29m 24s master
master
29m 24s
vulkan: workaround FA compile failures on macos (#13517) Release #43: Commit ab3971f pushed by ngxson
29m 39s master
master
29m 39s
quantize : improve tensor-type pattern matching (#13033) Release #42: Commit e5c834f pushed by ngxson
42m 38s master
master
42m 38s
clip : clip.h become private API (⚠️ breaking change) (#13510) Release #41: Commit 71bdbdb pushed by ngxson
51m 11s master
master
51m 11s
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
ProTip! You can narrow down the results and go further in time using created:<2025-05-12 or the other filters available.