From 781ddf786861f40de6d94d45d7b149d0f8d58c11 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Luka=20Govedi=C4=8D?= Date: Mon, 9 Feb 2026 10:05:14 -0500 Subject: [PATCH] [CI][torch.compile] Fix incorrect filtering for E2E fusion tests on B200 (#34031) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Luka Govedič --- .buildkite/test_areas/compile.yaml | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-) diff --git a/.buildkite/test_areas/compile.yaml b/.buildkite/test_areas/compile.yaml index 56fc011c77..51b9fdc8bb 100644 --- a/.buildkite/test_areas/compile.yaml +++ b/.buildkite/test_areas/compile.yaml @@ -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)"