Skip to content

Actions: ngxson/llama.cpp

Actions

Release

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
release.yml
995 workflow runs
995 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

context : remove redundant explicit casting to the same type (#15948) Release #745: Commit f4e664f pushed by ngxson
19m 36s master
master
19m 36s
server : adjust prompt similarity thold + add logs (#15913) Release #744: Commit f088b6a pushed by ngxson
20m 17s master
master
20m 17s
Vulkan iGPU device selection overhaul and PCI ID API support (#15947) Release #743: Commit 304ac56 pushed by ngxson
16m 17s master
master
16m 17s
vulkan: Make device memory check more portable (#15939) Release #742: Commit 6c88ad8 pushed by ngxson
19m 32s master
master
19m 32s
Revert "sycl: add usage of enqueue_functions extension (#14244)" (#1… Release #741: Commit 704d90c pushed by ngxson
23m 50s master
master
23m 50s
ggml-backend : add GGML_BACKEND_DEVICE_TYPE_IGPU device type (#15797) Release #740: Commit 360d653 pushed by ngxson
1h 2m 1s master
master
1h 2m 1s
CUDA: larger SRAM reads for tile FA, AMD FP16 dot (#15927) Release #739: Commit 0e6ff00 pushed by ngxson
50m 28s master
master
50m 28s
nitpick : correct MB to MiB (#15934) Release #738: Commit df082f5 pushed by ngxson
19m 12s master
master
19m 12s
ggml-cpu : add check for ARM MATMUL_INT8/i8mm support (#15922) Release #737: Commit 24a6734 pushed by ngxson
25m 39s master
master
25m 39s
kleidiai: fix GGML_ASSERT(*cur_backend_id != -1) failed (#15614) Release #736: Commit 2b3efea pushed by ngxson
21m 48s master
master
21m 48s
CANN: Disable acl_graph for prefill stage (#15933) Release #735: Commit c0389db pushed by ngxson
23m 30s master
master
23m 30s
CUDA: Add fastdiv to k_bin_bcast*, giving 1-3% E2E performance (#… Release #734: Commit 00681df pushed by ngxson
27m 42s master
master
27m 42s
llama : support T5 models with unequal number of encoder-decoder laye… Release #733: Commit 4f65885 pushed by ngxson
22m 9s master
master
22m 9s
graph : support non-contiguous Q in build_attn_mha (#15908) Release #732: Commit 6ab397e pushed by ngxson
28m 45s master
master
28m 45s
ggml-cpu : fix padding in ggml_timestep_embedding (#15917) Release #731: Commit 9de447d pushed by ngxson
18m 55s master
master
18m 55s
metal : make the backend async (#15906) Release #730: Commit 0f0a3c2 pushed by ngxson
13m 52s master
master
13m 52s
ci : add caching for ROCm installation in release workflow (#15924) Release #729: Commit 33daece pushed by ngxson
21m 5s master
master
21m 5s
tests : filter out no-ops from coverage report (#15900) Release #728: Commit e7b6d83 pushed by ngxson
24m 6s master
master
24m 6s
CANN: Add ROPE sin/cos cache for reuse (#15912) Release #727: Commit 10d8b2b pushed by ngxson
13m 53s master
master
13m 53s
CANN: implement LRU cache for ACL graphs (#15814) Release #726: Commit 28b5f19 pushed by ngxson
23m 51s master
master
23m 51s
llama : check returned fn ptrs from ggml_backend_reg_get_proc_address… Release #725: Commit 86587da pushed by ngxson
29m 36s master
master
29m 36s
vulkan: throw the oom error instead of no memory type found (#15905) Release #724: Commit ae355f6 pushed by ngxson
50m 48s master
master
50m 48s
vulkan: Fix OOB accesses in soft_max_back (#15861) Release #723: Commit 4f63cd7 pushed by ngxson
1h 9m 32s master
master
1h 9m 32s
HIP: use v_dot2_f32_f16 instruction for FA (#15884) Release #722: Commit 17bc5a8 pushed by ngxson
1h 19m 26s master
master
1h 19m 26s
Workaround for subgroup arithmetic failing on MoltenVK with AMD GPUs … Release #721: Commit ed54e32 pushed by ngxson
1h 7m 6s master
master
1h 7m 6s
ProTip! You can narrow down the results and go further in time using created:<2025-09-09 or the other filters available.