ci: Removing ROCm tests from trunk. (#164585)

Had a conversation with the AMD team today and I think we are all in
agreement that the current state of queueing for AMD is beyond where
we'd like to be for there to be blocking CI for ROCm.

Moving the representative testing jobs for this into the ciflow/rocm
workflow.

I'd love for these to be back in trunk if we can get to a state where
our queueing metrics are below an hour for ROCm infrastructure.

Dashboards:
* ROCm Queueing (>60mins) ([link](https://hud.pytorch.org/queue_time_analysis?dateRange=30&startDate=2025-09-03T16%3A06%3A45.025Z&endDate=2025-10-03T16%3A06%3A45.025Z&granularity=week&chartType=bar&repos=pytorch%2Fpytorch&category=machine_type&machineTypes=linux.rocm.gpu.2&machineTypes=linux.rocm.gpu.4&machineTypes=linux.rocm.gpu.mi250&machineTypes=linux.rocm.gpu.gfx942.1&machineTypes=linux.rocm.gpu.mi250.4&machineTypes=linux.rocm.gpu.gfx942.4&machineTypes=linux.rocm.gpu.mi355.2&machineTypes=linux.rocm.gpu.gfx942.4.test&machineTypes=linux.rocm.gpu.mi250.1&machineTypes=linux.rocm.gpu.gfx942.1.test&machineTypes=linux.rocm.gpu.gfx90a.1&machineTypes=linux.rocm.gpu.gfx90a.4&items=linux.rocm.gpu.2&items=linux.rocm.gpu.4&items=linux.rocm.gpu.mi250&items=linux.rocm.gpu.gfx942.1&items=linux.rocm.gpu.mi250.4&items=linux.rocm.gpu.gfx942.4&items=linux.rocm.gpu.mi355.2&items=linux.rocm.gpu.gfx942.4.test&items=linux.rocm.gpu.mi250.1&items=linux.rocm.gpu.gfx942.1.test&items=linux.rocm.gpu.gfx90a.1&items=linux.rocm.gpu.gfx90a.4))
* NVIDIA queueing (<5mins) ([link](https://hud.pytorch.org/queue_time_analysis?dateRange=30&startDate=2025-09-03T16%3A05%3A08.000Z&endDate=2025-10-03T16%3A05%3A08.000Z&granularity=week&chartType=bar&repos=pytorch%2Fpytorch&category=machine_type&machineTypes=lf.linux.g4dn.4xlarge.nvidia.gpu&machineTypes=linux.g4dn.12xlarge.nvidia.gpu&machineTypes=linux.g4dn.metal.nvidia.gpu&machineTypes=linux.g5.4xlarge.nvidia.gpu&machineTypes=lf.linux.g4dn.12xlarge.nvidia.gpu&machineTypes=lf.linux.g5.12xlarge.nvidia.gpu&machineTypes=lf.linux.g5.4xlarge.nvidia.gpu&machineTypes=lf.linux.g6.4xlarge.experimental.nvidia.gpu&machineTypes=linux.g6.4xlarge.experimental.nvidia.gpu&machineTypes=linux.4xlarge.nvidia.gpu&machineTypes=linux.g5.12xlarge.nvidia.gpu&machineTypes=linux.g4dn.4xlarge.nvidia.gpu&machineTypes=lf.linux.4xlarge.nvidia.gpu&machineTypes=linux.g6.12xlarge.nvidia.gpu&items=lf.linux.g4dn.4xlarge.nvidia.gpu&items=linux.g4dn.12xlarge.nvidia.gpu&items=linux.g4dn.metal.nvidia.gpu&items=linux.g5.4xlarge.nvidia.gpu&items=lf.linux.g4dn.12xlarge.nvidia.gpu&items=lf.linux.g5.12xlarge.nvidia.gpu&items=lf.linux.g5.4xlarge.nvidia.gpu&items=lf.linux.g6.4xlarge.experimental.nvidia.gpu&items=linux.g6.4xlarge.experimental.nvidia.gpu&items=linux.4xlarge.nvidia.gpu&items=linux.g5.12xlarge.nvidia.gpu&items=linux.g4dn.4xlarge.nvidia.gpu&items=lf.linux.4xlarge.nvidia.gpu&items=linux.g6.12xlarge.nvidia.gpu))

Signed-off-by: Eli Uriegas <eliuriegas@meta.com>

Pull Request resolved: https://github.com/pytorch/pytorch/pull/164585
Approved by: https://github.com/malfet, https://github.com/yangw-dev, https://github.com/atalman, https://github.com/jeffdaily
This commit is contained in:
Eli Uriegas
2025-10-03 10:47:16 -07:00
committed by PyTorch MergeBot
parent f39789cdab
commit 7eb1eb4313

View File

@ -189,41 +189,6 @@ jobs:
runner: "${{ needs.get-label-type.outputs.label-type }}windows.4xlarge.nonephemeral"
secrets: inherit
linux-jammy-rocm-py3_10-build:
if: ${{ startsWith(github.event.ref, 'refs/tags/ciflow/trunk') }}
name: linux-jammy-rocm-py3.10
uses: ./.github/workflows/_linux-build.yml
needs: get-label-type
with:
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
build-environment: linux-jammy-rocm-py3.10
docker-image-name: ci-image:pytorch-linux-jammy-rocm-n-py3
sync-tag: rocm-build
test-matrix: |
{ include: [
{ config: "default", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
{ config: "default", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.gfx942.1" },
{ config: "distributed", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.gfx942.4" },
]}
secrets: inherit
linux-jammy-rocm-py3_10-test:
if: ${{ startsWith(github.event.ref, 'refs/tags/ciflow/trunk') }}
permissions:
id-token: write
contents: read
name: linux-jammy-rocm-py3.10
uses: ./.github/workflows/_rocm-test.yml
needs:
- linux-jammy-rocm-py3_10-build
- target-determination
with:
build-environment: linux-jammy-rocm-py3.10
docker-image: ${{ needs.linux-jammy-rocm-py3_10-build.outputs.docker-image }}
test-matrix: ${{ needs.linux-jammy-rocm-py3_10-build.outputs.test-matrix }}
tests-to-include: "test_nn test_torch test_cuda test_ops test_unary_ufuncs test_binary_ufuncs test_autograd inductor/test_torchinductor distributed/test_c10d_common distributed/test_c10d_nccl"
secrets: inherit
inductor-build:
name: inductor-build
uses: ./.github/workflows/_linux-build.yml