biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[CI][torch.compile] Fix incorrect filtering for E2E fusion tests on B200 (#34031)

Browse Source 
Signed-off-by: Luka Govedič <lgovedic@redhat.com>
This commit is contained in:
Luka Govedič
2026-02-09 10:05:14 -05:00
committed by GitHub
parent 64a9c2528b
commit 781ddf7868
1 changed files with 8 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
.buildkite/test_areas/compile.yaml
View File

@@ -121,13 +121,10 @@ steps:
optional: true
commands:
- nvidia-smi
# Run all models and attn backends but only Inductor partition and native custom ops
# -k "inductor_partition and not +rms_norm and not +quant_fp8"
# Run all models but only FLASHINFER, Inductor partition and native custom ops
# Qwen requires +quant_fp8 as -quant_fp8 rms+quant fusion is not supported
# -k "inductor_partition and not +rms_norm and +quant_fp8 and qwen3"
# Run just llama3 (fp8 & fp4) for all config combinations
# -k "llama-3"
- pytest -v -s tests/compile/fusions_e2e/test_tp1_quant.py -k "inductor_partition and not +rms_norm and not +quant_fp8" -k "inductor_partition and not +rms_norm and +quant_fp8 and qwen3" -k "llama-3"
# Run just llama3 (fp8 & fp4) for all config combinations (only inductor partition)
- pytest -v -s tests/compile/fusions_e2e/test_tp1_quant.py -k "inductor_partition and (FLASHINFER and not +rms_norm and (not +quant_fp8 or +quant_fp8 and qwen3) or llama-3)"
- label: Fusion E2E TP2 Quick (H100)
timeout_in_minutes: 20
@@ -162,7 +159,7 @@ steps:
- tests/compile/fusions_e2e/
commands:
- nvidia-smi
# Run just llama3 (fp4 & fp8 & bf16) for all config combinations
# Run just llama3 (fp8 & bf16) for all config combinations
- pytest -v -s tests/compile/fusions_e2e/test_tp2_ar_rms.py -k "llama-3"
- label: Fusion E2E TP2 AsyncTP Config Sweep (H100)
@@ -197,7 +194,8 @@ steps:
- tests/compile/fusions_e2e/
commands:
- nvidia-smi
# Run all models and attn backends but only Inductor partition and native custom ops
# Run all models but only FLASHINFER, Inductor partition and native custom ops
# include qwen with +quant_fp8 as -quant_fp8 rms+quant fusion is not supported
# for ar-rms-quant-fp4, also sweep llama3
- pytest -v -s tests/compile/fusions_e2e/test_tp2_ar_rms.py -k "inductor_partition and not +rms_norm and not +quant_fp8" -k "Llama-3.1-8B-Instruct-FP4"
- pytest -v -s tests/compile/fusions_e2e/test_tp2_async_tp.py -k "inductor_partition and not +rms_norm and not +quant_fp8"
- pytest -v -s tests/compile/fusions_e2e/test_tp2_ar_rms.py -k "(FLASHINFER and inductor_partition and not +rms_norm and (not +quant_fp8 or +quant_fp8 and qwen3)) or Llama-3.1-8B-Instruct-FP4"
- pytest -v -s tests/compile/fusions_e2e/test_tp2_async_tp.py -k "FLASHINFER and inductor_partition and not +rms_norm and (not +quant_fp8 or +quant_fp8 and qwen3)"