Skip to content

vulkan: fix fp16 Flash Attention on Windows AMD RDNA2 and below (#19921) #42

vulkan: fix fp16 Flash Attention on Windows AMD RDNA2 and below (#19921)

vulkan: fix fp16 Flash Attention on Windows AMD RDNA2 and below (#19921) #42

Triggered via push February 26, 2026 19:44
Status Cancelled
Total duration 1d 0h 0m 6s
Artifacts 20

release.yml

on: push
Matrix: openEuler-cann
Matrix: ubuntu-22-cpu
Matrix: ubuntu-22-rocm
Matrix: windows-cpu
Matrix: windows-cuda
Matrix: windows-hip
Matrix: windows
release
0s
release
Fit to window
Zoom out
Zoom in

Annotations

1 error and 4 warnings
ubuntu-22-cpu (s390x, ubuntu-24.04-s390x)
The job has exceeded the maximum execution time while awaiting a runner for 24h0m0s
windows (vulkan, x64, -DGGML_VULKAN=ON, ggml-vulkan)
Cache not found for keys: ccache-windows-latest-cmake-vulkan-x64-
openEuler-cann (aarch64, 910b, Release, on)
The command [sudo apt-get remove -y azure-cli google-chrome-stable firefox powershell mono-devel libgl1-mesa-dri --fix-missing] failed to complete successfully. Proceeding...
openEuler-cann (aarch64, 310p, Release, off)
The command [sudo apt-get remove -y azure-cli google-chrome-stable firefox powershell mono-devel libgl1-mesa-dri --fix-missing] failed to complete successfully. Proceeding...
windows-hip (radeon, gfx1150;gfx1151;gfx1200;gfx1201;gfx1100;gfx1101;gfx1102;gfx1030;gfx1031;gfx1...
Cache not found for keys: ccache-windows-latest-cmake-hip-26.Q1-radeon-x64-

Artifacts

Produced during runtime
Name Size Digest
cudart-llama-bin-win-cuda-12.4-x64.zip
372 MB
sha256:ab54562ad51237e94f4a6050f169e19733ea3683427b98b1856c255fb70562d5
cudart-llama-bin-win-cuda-13.1-x64.zip
383 MB
sha256:c3d1e968a3d79cfabd82e984bebce0ba6b55812671d1b1c4a26d08fcf50c2ef4
llama-b8168-xcframework.zip
160 MB
sha256:1645f3a4b22f9c1e366beeec9dea008f449303c45afa5b0582fc4b09a214b8bd
llama-bin-310p-openEuler-aarch64.tar.gz
53.6 MB
sha256:604d54de7ae30ed471df21b922b1525aaa195236ff699fa22822437e3372c6d8
llama-bin-310p-openEuler-x86.tar.gz
59.4 MB
sha256:0bffdd155326d751857007625f19d45d39f668fa1a55f5644eda9c3da7ce669c
llama-bin-910b-openEuler-aarch64-aclgraph.tar.gz
53.6 MB
sha256:69428dcee913f2347cc0a66f245cfd9a90ed5e6408c42dbbc8b73a41523fc757
llama-bin-910b-openEuler-x86-aclgraph.tar.gz
59.4 MB
sha256:10f19d0cbafadc8cee55910bd45ced69254b74765568aa7cffc4289af12277ba
llama-bin-macos-arm64.tar.gz
29.1 MB
sha256:6d6b4e041f877bb319c6ae1613b151bd2e687570b4b8cea7586e127bca5c7bab
llama-bin-macos-x64.tar.gz
82.5 MB
sha256:245166ad7840f8ed0e8e32cd9fae3ceef65ba7783e03f2c659af06c8ad9ccd9e
llama-bin-ubuntu-rocm-7.2-x64.tar.gz
129 MB
sha256:93c09ab90885d34b2b48176931a7561f508294450c15bdf3221b6d94b7831cd7
llama-bin-ubuntu-vulkan-x64.tar.gz
39.5 MB
sha256:5393ac659be31ce9fdc0a29acea0dbbafe35d30b09632843eb5e7aa5bdd73d28
llama-bin-ubuntu-x64.tar.gz
23.7 MB
sha256:605a804581ed575605ca73ffabeb07e8283c475014adbf286c8190d0d27ce2ae
llama-bin-win-cpu-arm64.zip
23.5 MB
sha256:e230e51570895c6de8bc16f8c4e2d7b2ae4804b4939e277efe97cbacd35e7cec
llama-bin-win-cpu-x64.zip
29.7 MB
sha256:1e637af636ea3c38e1d7b93f502d2910574b5e05b32150a77d60e8f73d389c44
llama-bin-win-cuda-12.4-x64.zip
178 MB
sha256:0f2e2f66603f14411a5430ca38992b37f213fe2a4ecf931333a7aae6e696b8d5
llama-bin-win-cuda-13.1-x64.zip
111 MB
sha256:f27643c7d70e131a629b4c364a3ec77b8f33405ef9a4463a05627d22715a1640
llama-bin-win-hip-radeon-x64.zip
295 MB
sha256:210037aeea8da4a52a8b79272b1fb3e10ddb8ecf8a8c2eadef64cbc530e06d81
llama-bin-win-opencl-adreno-arm64.zip
150 KB
sha256:8bf687d810567a526f7fdbc67ad5ecba70797d563d11a357603c82515fce2501
llama-bin-win-sycl-x64.zip
84.1 MB
sha256:929f05c625b519bbef4fafb64bca9e1e9a5d083724d39144277957c05b0af82d
llama-bin-win-vulkan-x64.zip
14.5 MB
sha256:5c7197b886309ec6e277f633a40b64fa55f848980ae3fa17fa4fcf868df58e50