[AMD][OCP MX][CI] Fix tests to not dispatch on UNFUSED_TRITON backend on MI300, improve w_mxfp4_a_fp8 emulation support (#46142)

Signed-off-by: Felix Marty <Felix.Marty@amd.com>
This commit is contained in:
fxmarty-amd
2026-06-23 14:25:27 -04:00
committed by GitHub
parent ceae5bcbda
commit e368415daa
7 changed files with 74 additions and 17 deletions
@@ -104,6 +104,11 @@ def test_gpt_oss_attention_quantization(
model_args = EvaluationConfig(model_name).get_model_args(tp_size)
# Emulation backend on MI300, MI250 is opt-in
# following https://github.com/vllm-project/vllm/pull/45896
if not on_gfx950():
model_args["moe_backend"] = "emulation"
extra_run_kwargs = {
"gen_kwargs": {"max_gen_toks": 8000},
"apply_chat_template": True,