ci: dynamic partition + LPT from live sglang-ci-stats model (#25263)
This commit is contained in:
@@ -372,6 +372,7 @@ jobs:
|
||||
check_changes: ${{ toJson(needs.check-changes.outputs) }}
|
||||
caller_inputs: ${{ toJson(inputs) }}
|
||||
partitions: ${{ needs.check-changes.outputs.partitions }}
|
||||
run_timeout_minutes: '30'
|
||||
secrets: inherit
|
||||
|
||||
# Runs on H100 (80GB, SM90) - tests that don't pass on 5090 (FA3, FP8, high VRAM, etc.)
|
||||
@@ -386,6 +387,7 @@ jobs:
|
||||
check_changes: ${{ toJson(needs.check-changes.outputs) }}
|
||||
caller_inputs: ${{ toJson(inputs) }}
|
||||
partitions: ${{ needs.check-changes.outputs.partitions }}
|
||||
run_timeout_minutes: '30'
|
||||
timeout_per_file: '1800'
|
||||
secrets: inherit
|
||||
|
||||
@@ -400,6 +402,7 @@ jobs:
|
||||
check_changes: ${{ toJson(needs.check-changes.outputs) }}
|
||||
caller_inputs: ${{ toJson(inputs) }}
|
||||
partitions: ${{ needs.check-changes.outputs.partitions }}
|
||||
run_timeout_minutes: '30'
|
||||
secrets: inherit
|
||||
|
||||
stage-b-test-4-gpu-b200:
|
||||
@@ -461,6 +464,7 @@ jobs:
|
||||
check_changes: ${{ toJson(needs.check-changes.outputs) }}
|
||||
caller_inputs: ${{ toJson(inputs) }}
|
||||
partitions: ${{ needs.check-changes.outputs.partitions }}
|
||||
run_timeout_minutes: '30'
|
||||
secrets: inherit
|
||||
|
||||
stage-c-test-8-gpu-h200:
|
||||
@@ -474,6 +478,7 @@ jobs:
|
||||
check_changes: ${{ toJson(needs.check-changes.outputs) }}
|
||||
caller_inputs: ${{ toJson(inputs) }}
|
||||
partitions: ${{ needs.check-changes.outputs.partitions }}
|
||||
run_timeout_minutes: '30'
|
||||
# Per-model TP must match the test's launch in test/registered/ -- see
|
||||
# FALLBACK_ARGS in scripts/ci/cuda/warmup_deep_gemm.py for extra dp/ep
|
||||
# flags. Only models that actually invoke DeepGEMM kernels at runtime
|
||||
@@ -494,6 +499,7 @@ jobs:
|
||||
check_changes: ${{ toJson(needs.check-changes.outputs) }}
|
||||
caller_inputs: ${{ toJson(inputs) }}
|
||||
partitions: ${{ needs.check-changes.outputs.partitions }}
|
||||
run_timeout_minutes: '30'
|
||||
secrets: inherit
|
||||
|
||||
stage-c-test-deepep-4-gpu-h100:
|
||||
@@ -507,6 +513,7 @@ jobs:
|
||||
check_changes: ${{ toJson(needs.check-changes.outputs) }}
|
||||
caller_inputs: ${{ toJson(inputs) }}
|
||||
partitions: ${{ needs.check-changes.outputs.partitions }}
|
||||
run_timeout_minutes: '30'
|
||||
warmup_deep_gemm_models: 'lmsys/sglang-ci-dsv3-test:4'
|
||||
warmup_server_models: 'lmsys/sglang-ci-dsv3-test:4'
|
||||
secrets: inherit
|
||||
@@ -538,6 +545,7 @@ jobs:
|
||||
check_changes: ${{ toJson(needs.check-changes.outputs) }}
|
||||
caller_inputs: ${{ toJson(inputs) }}
|
||||
partitions: ${{ needs.check-changes.outputs.partitions }}
|
||||
run_timeout_minutes: '30'
|
||||
timeout_per_file: '1800'
|
||||
secrets: inherit
|
||||
|
||||
@@ -552,6 +560,7 @@ jobs:
|
||||
check_changes: ${{ toJson(needs.check-changes.outputs) }}
|
||||
caller_inputs: ${{ toJson(inputs) }}
|
||||
partitions: ${{ needs.check-changes.outputs.partitions }}
|
||||
run_timeout_minutes: '45'
|
||||
timeout_per_file: '1800'
|
||||
secrets: inherit
|
||||
|
||||
@@ -566,6 +575,7 @@ jobs:
|
||||
check_changes: ${{ toJson(needs.check-changes.outputs) }}
|
||||
caller_inputs: ${{ toJson(inputs) }}
|
||||
partitions: ${{ needs.check-changes.outputs.partitions }}
|
||||
run_timeout_minutes: '30'
|
||||
timeout_per_file: '1800'
|
||||
secrets: inherit
|
||||
|
||||
|
||||
Reference in New Issue
Block a user