Enabling some B200-specific tests on MI355 (#35253)

Signed-off-by: Alexei V. Ivanov <alexei.ivanov@amd.com>
Signed-off-by: Alexei-V-Ivanov-AMD <156011006+Alexei-V-Ivanov-AMD@users.noreply.github.com>
This commit is contained in:
Alexei-V-Ivanov-AMD
2026-03-06 19:27:20 +00:00
committed by GitHub
parent f3c6c9c9d7
commit 225d1090a0
4 changed files with 89 additions and 110 deletions
@@ -0,0 +1,9 @@
model_name: "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8"
accuracy_threshold: 0.85
num_questions: 1319
num_fewshot: 5
server_args: >-
--max-model-len 4096
--tensor-parallel-size 2
--enable-expert-parallel
--async-scheduling
@@ -0,0 +1,5 @@
Qwen3-0.6B-FP8.yaml
Qwen2.5-VL-3B-Instruct-FP8-dynamic.yaml
Qwen1.5-MoE-W4A16-CT.yaml
DeepSeek-V2-Lite-Instruct-FP8.yaml
Qwen3-Next-FP8-EP2_MI355.yaml
@@ -343,6 +343,10 @@ def test_auto_backend_selection_behavior():
("FLEX_ATTENTION", None, False), # Flex does not support
],
)
@pytest.mark.skipif(
current_platform.is_rocm(),
reason="Attention backend FA3 is not supported on ROCm. This test can't succeed.",
)
def test_per_head_quant_scales_backend_selection(
backend_name: str, flash_attn_version: int | None, should_succeed: bool
):