Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

Revert "ignore mtp.fc for qwen3_5 due to vllm failure (#1710)" (#1730)

Xin Hec2344554d863
diff --git a/auto_round/special_model_handler.py b/auto_round/special_model_handler.py
index c07b07d5e..1a336ee6f 100644
--- a/auto_round/special_model_handler.py
+++ b/auto_round/special_model_handler.py
@@ -719,18 +719,6 @@ register_ignore_layers(
],
)
-register_ignore_layers(
- matchers=[
- ModelTypeMatcher(r"qwen3_5", mode="in"),
- ],
- ignore_layers=[
- "mlp.gate",
- "mlp.shared_expert_gate",
- "mtp.fc",
- "mtp.gate",
- ],
-)
-
def get_predefined_ignore_layers(model: torch.nn.Module) -> list[str]:
layers = []