[ROCm][CI] Expand AITER W4A4 MoE Coverage (#52647)

Signed-off-by: Micah Williamson <micah.williamson@amd.com>
This commit is contained in:
Micah Williamson
2026-08-17 15:46:22 -05:00
committed by GitHub
parent f08a95f8d8
commit 8878ebd8fd
2 changed files with 59 additions and 7 deletions
+23 -2
View File
@@ -125,7 +125,6 @@ def test_mxfp4_loading_and_execution_moe(vllm_runner, model_case: ModelCase):
# if model_case.model_id == "fxmarty/qwen_1.5-moe-a2.7b-mxfp4":
# llm.apply_model(check_model)
output = llm.generate_greedy("Today I am in the French Alps and", max_tokens=20)
assert output
@@ -187,6 +186,16 @@ def mxfp8_dequantize(x, scale):
return x_float * scale
def mxfp4_quant_dequant(x: torch.Tensor) -> torch.Tensor:
shape = x.shape
quantized, scale = dynamic_mxfp4_quant(x.to(torch.bfloat16).flatten(0, -2))
return (
upcast_from_mxfp(quantized.view(torch.uint8), scale, torch.bfloat16, axis=-1)
.reshape(shape)
.float()
)
def reference_moe(
roouting_logits,
topk,
@@ -219,6 +228,8 @@ def reference_moe(
expert_weights = torch.nn.functional.softmax(experts.values, dim=1)
expert_indices = experts.indices
t = hidden_states.clone()
if act_type == "mxfp4":
t = mxfp4_quant_dequant(t)
# MLP #1
mlp1_weight = w13[expert_indices, ...]
mlp1_bias = bias13[expert_indices, ...]
@@ -244,6 +255,8 @@ def reference_moe(
t.to(torch.bfloat16), is_sf_swizzled_layout=False
)
t = mxfp8_dequantize(t_quantized, t_scale)
elif act_type == "mxfp4":
t = mxfp4_quant_dequant(t)
# MLP #2
mlp2_weight = w2[expert_indices, ...]
mlp2_bias = bias2[expert_indices, ...]
@@ -1392,6 +1405,14 @@ ROCM_BACKEND_CONFIGS = {
"requires_aiter": True,
"requires_gfx950": True,
},
"AITER_MXFP4_MXFP4": {
"activation": "SILU",
"act_type": "mxfp4",
"rtol": 1.0,
"percent": 0.8,
"requires_aiter": True,
"requires_gfx950": True,
},
}
@@ -1662,7 +1683,7 @@ def test_rocm_mxfp4_moe_oracle(
alpha=1.702 if activation == MoEActivation.SWIGLUOAI else 1.0,
beta=1.0 if activation == MoEActivation.SWIGLUOAI else 0.0,
limit=7.0 if activation == MoEActivation.SWIGLUOAI else None,
act_type="bf16",
act_type=str(config.get("act_type", "bf16")),
activation=act_name,
use_interleaved_layout=use_interleaved,
)