Skip to content

Lower grouped GEMM on sm_80 and above (#25017) #22429

Lower grouped GEMM on sm_80 and above (#25017)

Lower grouped GEMM on sm_80 and above (#25017) #22429

Triggered via pull request November 13, 2026 04:32
Status Cancelled
Total duration 4h 8m 56s
Artifacts 2

cuda.yml

on: pull_request
Get changed files  /  get-changed-files
4s
Get changed files / get-changed-files
CI run decision  /  decide
34s
CI run decision / decide
Validate benchmark dispatch inputs
3s
Validate benchmark dispatch inputs
Matrix: test-cuda-builds
export-cuda-multi-arch-a100  /  linux-job
17m 6s
export-cuda-multi-arch-a100 / linux-job
test-models-cuda  /  linux-job
1h 12m
test-models-cuda / linux-job
unittest-cuda-muse-glimmer-batching  /  linux-job
53m 49s
unittest-cuda-muse-glimmer-batching / linux-job
unittest-cuda-runtime  /  linux-job
26m 15s
unittest-cuda-runtime / linux-job
Matrix: test-muse-glimmer-cuda-e2e
Matrix: unittest-cuda-quantized-gemm
Matrix: unittest-cuda
Matrix: test-model-cuda-e2e
Waiting for pending jobs
check-all-cuda-builds
4s
check-all-cuda-builds
merge-and-test-cuda-multi-arch-a10g  /  linux-job
23m 54s
merge-and-test-cuda-multi-arch-a10g / linux-job
upload-benchmark-results
0s
upload-benchmark-results
test-cuda-merged-pte-a100  /  linux-job
16m 34s
test-cuda-merged-pte-a100 / linux-job
Fit to window
Zoom out
Zoom in

Annotations

64 errors, 16 warnings, and 4 notices
export-cuda-multi-arch-a100 / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
export-cuda-multi-arch-a100 / linux-job
Process completed with exit code 1.
export-cuda-multi-arch-a100 / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
export-cuda-multi-arch-a100 / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
test-executorch-cuda-build-13.0 / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
test-executorch-cuda-build-13.0 / linux-job
Process completed with exit code 1.
test-executorch-cuda-build-13.0 / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
test-executorch-cuda-build-13.0 / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
test-executorch-cuda-build-13.4 / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
test-executorch-cuda-build-13.4 / linux-job
Process completed with exit code 1.
test-executorch-cuda-build-13.4 / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
test-executorch-cuda-build-13.4 / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
unittest-cuda-runtime / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
unittest-cuda-runtime / linux-job
Process completed with exit code 1.
unittest-cuda-runtime / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
unittest-cuda-runtime / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
merge-and-test-cuda-multi-arch-a10g / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
merge-and-test-cuda-multi-arch-a10g / linux-job
Process completed with exit code 1.
merge-and-test-cuda-multi-arch-a10g / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
merge-and-test-cuda-multi-arch-a10g / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
unittest-cuda-quantized-gemm (int4) / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
unittest-cuda-quantized-gemm (int4) / linux-job
Process completed with exit code 1.
unittest-cuda-quantized-gemm (int4) / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
unittest-cuda-quantized-gemm (int4) / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
unittest-cuda-muse-glimmer-batching / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
unittest-cuda-muse-glimmer-batching / linux-job
Process completed with exit code 1.
unittest-cuda-muse-glimmer-batching / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
unittest-cuda-muse-glimmer-batching / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
unittest-cuda-quantized-gemm (large-int5-int6-int8-graph-timing) / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
unittest-cuda-quantized-gemm (large-int5-int6-int8-graph-timing) / linux-job
Process completed with exit code 1.
unittest-cuda-quantized-gemm (large-int5-int6-int8-graph-timing) / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
unittest-cuda-quantized-gemm (large-int5-int6-int8-graph-timing) / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
test-muse-glimmer-cuda-e2e-kquant-dynamic-solo-text / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
test-muse-glimmer-cuda-e2e-kquant-dynamic-solo-text / linux-job
Process completed with exit code 1.
test-muse-glimmer-cuda-e2e-kquant-dynamic-solo-text / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
test-muse-glimmer-cuda-e2e-kquant-dynamic-solo-text / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
unittest-cuda (system) / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
unittest-cuda (system) / linux-job
Process completed with exit code 1.
unittest-cuda (system) / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
unittest-cuda (system) / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
unittest-cuda (kernel) / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
unittest-cuda (kernel) / linux-job
Process completed with exit code 1.
unittest-cuda (kernel) / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
unittest-cuda (kernel) / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
test-models-cuda / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
test-models-cuda / linux-job
Process completed with exit code 1.
test-models-cuda / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
test-models-cuda / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
test-cuda-merged-pte-a100 / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
test-cuda-merged-pte-a100 / linux-job
Process completed with exit code 1.
test-cuda-merged-pte-a100 / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
test-cuda-merged-pte-a100 / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
test-muse-glimmer-cuda-e2e-kquant-dynamic-dflash-image / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
test-muse-glimmer-cuda-e2e-kquant-dynamic-dflash-image / linux-job
Process completed with exit code 1.
test-muse-glimmer-cuda-e2e-kquant-dynamic-dflash-image / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
test-muse-glimmer-cuda-e2e-kquant-dynamic-dflash-image / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
test-muse-glimmer-cuda-e2e-kquant-17gb-dflash-image-batching / linux-job
The job has exceeded the maximum execution time of 4h0m0s
test-muse-glimmer-cuda-e2e-kquant-17gb-dflash-image-batching / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
test-muse-glimmer-cuda-e2e-kquant-17gb-dflash-image-batching / linux-job
[OSDC] Step cancelled by GitHub Actions (received SIGINT). This typically means matrix fail-fast, a manual cancel, a workflow concurrency override, or a run-level cancellation — NOT an OSDC infrastructure failure. Check the workflow-run page for the actual cause; the in-pod subprocess is terminated by the hook so post-cleanup steps can run.
test-muse-glimmer-cuda-e2e-kquant-17gb-dflash-image-batching / linux-job
Executing the custom container implementation failed. Please contact your self hosted runner administrator.
test-muse-glimmer-cuda-e2e-kquant-17gb-dflash-image-batching / linux-job
Process completed with exit code 1.
test-muse-glimmer-cuda-e2e-kquant-17gb-dflash-image-batching / linux-job
[OSDC] Step script exited with code 1. This is a script/workflow error, not an infrastructure issue. Check the step logs above for the actual failure.
test-muse-glimmer-cuda-e2e-kquant-17gb-dflash-image-batching / linux-job
Credentials could not be loaded, please check your action inputs: Could not load credentials from any providers
CI run decision / decide
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@v4. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
export-cuda-multi-arch-a100 / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
test-executorch-cuda-build-13.0 / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
test-executorch-cuda-build-13.4 / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
unittest-cuda-runtime / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
merge-and-test-cuda-multi-arch-a10g / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093, actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
unittest-cuda-quantized-gemm (int4) / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
unittest-cuda-muse-glimmer-batching / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
unittest-cuda-quantized-gemm (large-int5-int6-int8-graph-timing) / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
test-muse-glimmer-cuda-e2e-kquant-dynamic-solo-text / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
unittest-cuda (system) / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
unittest-cuda (kernel) / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
test-models-cuda / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
test-cuda-merged-pte-a100 / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
test-muse-glimmer-cuda-e2e-kquant-dynamic-dflash-image / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
test-muse-glimmer-cuda-e2e-kquant-17gb-dflash-image-batching / linux-job
Node.js 20 is deprecated. The following actions target Node.js 20 but are being forced to run on Node.js 24: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683, aws-actions/configure-aws-credentials@67fbcbb121271f7775d2e7715933280b06314838, pmeier/pytest-results-action@a2c1430e2bddadbad9f49a6f9b879f062c6b19b1. For more information see: https://github.blog/changelog/2025-09-19-deprecation-of-node-20-on-github-actions-runners/
Validate benchmark dispatch inputs
"The ubuntu-latest label will migrate to Ubuntu 26 beginning October 19, 2026. For more information, see https://github.com/actions/runner-images/issues/14748"
Get changed files / get-changed-files
"The ubuntu-latest label will migrate to Ubuntu 26 beginning October 19, 2026. For more information, see https://github.com/actions/runner-images/issues/14748"
CI run decision / decide
"The ubuntu-latest label will migrate to Ubuntu 26 beginning October 19, 2026. For more information, see https://github.com/actions/runner-images/issues/14748"
check-all-cuda-builds
"The ubuntu-latest label will migrate to Ubuntu 26 beginning October 19, 2026. For more information, see https://github.com/actions/runner-images/issues/14748"

Artifacts

Produced during runtime
Name Size Digest
cuda-multi-arch-a100
1.28 MB
sha256:90b9ebbbf4df6fdd721fe593778eb60aeafc6e435451ba07f2368eb20bc8dd99
cuda-multi-arch-merged
2.96 MB
sha256:f5447a36f4f18e8298c3cf804da3a1848764060742c34e8710003b18f50b753b