From acd2662121006a8df91996d5bea4e0f9e2985a5e Mon Sep 17 00:00:00 2001 From: jimmzhou Date: Fri, 13 Mar 2026 16:27:13 +0000 Subject: [PATCH 1/3] skip sliding window --- tests/attention/test_fmha_v2_prefill.py | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/tests/attention/test_fmha_v2_prefill.py b/tests/attention/test_fmha_v2_prefill.py index c31aeb2b94..d5dd382d14 100644 --- a/tests/attention/test_fmha_v2_prefill.py +++ b/tests/attention/test_fmha_v2_prefill.py @@ -837,6 +837,8 @@ def test_trtllm_fmha_v2_prefill( and mask_mode == "SLIDING_WINDOW" ): pytest.skip("Skip due to bug in fp8 sliding window") + if mask_mode == "SLIDING_WINDOW": + pytest.skip("todo(jimmyzho): temporarily skip sliding window test due to hang") run_trtllm_fmha_v2_prefill_case( input_layout=input_layout, batch_size=batch_size, @@ -955,7 +957,8 @@ def test_trtllm_fmha_v2_prefill_attention_sinks( if not is_sm90a_supported(torch.device("cuda")): pytest.skip("FMHA v2 requires SM90+ (Hopper) GPUs.") - + if mask_mode == "SLIDING_WINDOW": + pytest.skip("todo(jimmyzho): temporarily skip sliding window test due to hang") torch.manual_seed(42) device = torch.device("cuda") From 6888c0b54b2da75f5655eb29517daa60cd09e861 Mon Sep 17 00:00:00 2001 From: jimmzhou Date: Fri, 13 Mar 2026 16:43:48 +0000 Subject: [PATCH 2/3] skip fp8 --- tests/attention/test_fmha_v2_prefill.py | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tests/attention/test_fmha_v2_prefill.py b/tests/attention/test_fmha_v2_prefill.py index d5dd382d14..be6673dac8 100644 --- a/tests/attention/test_fmha_v2_prefill.py +++ b/tests/attention/test_fmha_v2_prefill.py @@ -839,6 +839,8 @@ def test_trtllm_fmha_v2_prefill( pytest.skip("Skip due to bug in fp8 sliding window") if mask_mode == "SLIDING_WINDOW": pytest.skip("todo(jimmyzho): temporarily skip sliding window test due to hang") + if dtype == torch.float8_e4m3fn and o_dtype == torch.float8_e4m3fn: + pytest.skip("todo(jimmyzho): temporarily skip fp8 tests due to hang") run_trtllm_fmha_v2_prefill_case( input_layout=input_layout, batch_size=batch_size, From 38a1003dd816173686884e235863f243a6666995 Mon Sep 17 00:00:00 2001 From: Brian Ryu Date: Fri, 13 Mar 2026 19:25:15 +0000 Subject: [PATCH 3/3] Skip entire file --- tests/attention/test_fmha_v2_prefill.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/tests/attention/test_fmha_v2_prefill.py b/tests/attention/test_fmha_v2_prefill.py index be6673dac8..90d075c649 100644 --- a/tests/attention/test_fmha_v2_prefill.py +++ b/tests/attention/test_fmha_v2_prefill.py @@ -3,6 +3,10 @@ import math from typing import Optional, Tuple, Union +pytestmark = pytest.mark.skip( + reason="todo(jimmyzho): temporarily skip this test due to hangs" +) + import flashinfer from flashinfer.prefill import fmha_v2_prefill_deepseek from tests.utils_fp8 import to_float8