[CI] Give the kernel lane a 5090 suite and move kernel-only tests off the general lane (#40496)
This commit is contained in:
@@ -42,7 +42,9 @@ env:
|
||||
jobs:
|
||||
# `name:` reproduces the job names this workflow published before the table.
|
||||
jit-kernel-test:
|
||||
name: ${{ matrix.job_name }}
|
||||
# The job name is the suite it runs, with the stage prefix swapped for this
|
||||
# workflow's, so a row declares a (suite, runner) pair and nothing else.
|
||||
name: jit-kernel-${{ matrix.suite }}${{ matrix.shard }}
|
||||
# Runs whenever call-jit-kernel-tests dispatches this workflow. That caller is the
|
||||
# single gate (PR jit_kernel changes, or scheduled/parallel-dispatch full runs), so
|
||||
# the sub-jobs no longer re-exclude schedule/parallel-dispatch here.
|
||||
@@ -50,25 +52,24 @@ jobs:
|
||||
fail-fast: false
|
||||
matrix:
|
||||
include:
|
||||
- job_name: jit-kernel-unit-test (0)
|
||||
- suite: unit-test-1-gpu-large
|
||||
runner_config: 1-gpu-large
|
||||
suite: base-b-kernel-unit-test-1-gpu-large
|
||||
shard: " (0)"
|
||||
suite_args: --auto-partition-id 0 --auto-partition-size 2 --fork-worker-batch-size 20
|
||||
- job_name: jit-kernel-unit-test (1)
|
||||
- suite: unit-test-1-gpu-large
|
||||
runner_config: 1-gpu-large
|
||||
suite: base-b-kernel-unit-test-1-gpu-large
|
||||
shard: " (1)"
|
||||
suite_args: --auto-partition-id 1 --auto-partition-size 2 --fork-worker-batch-size 20
|
||||
- job_name: jit-kernel-multigpu-unit-test
|
||||
- suite: unit-test-1-gpu-small
|
||||
runner_config: 1-gpu-small
|
||||
- suite: unit-test-4-gpu-b200
|
||||
runner_config: 4-gpu-b200
|
||||
- suite: unit-test-8-gpu-h200
|
||||
runner_config: 8-gpu-h200
|
||||
suite: base-b-kernel-unit-test-8-gpu-h200
|
||||
# Alone among these, it has never run the health check.
|
||||
skip_health_check: true
|
||||
- job_name: jit-kernel-benchmark-test
|
||||
- suite: benchmark-test-1-gpu-large
|
||||
runner_config: 1-gpu-large
|
||||
suite: base-b-kernel-benchmark-test-1-gpu-large
|
||||
- job_name: jit-kernel-b200-test
|
||||
runner_config: 4-gpu-b200
|
||||
suite: base-b-kernel-unit-test-4-gpu-b200
|
||||
runs-on: ${{ fromJson(inputs.runs_on_map)[matrix.runner_config] }}
|
||||
timeout-minutes: 240
|
||||
steps:
|
||||
@@ -108,4 +109,4 @@ jobs:
|
||||
timeout-minutes: 60
|
||||
run: |
|
||||
cd test/
|
||||
python3 run_suite.py --hw cuda --suite ${{ matrix.suite }} ${{ matrix.suite_args }}
|
||||
python3 run_suite.py --hw cuda --suite base-b-kernel-${{ matrix.suite }} ${{ matrix.suite_args }}
|
||||
|
||||
Reference in New Issue
Block a user