diff --git a/python/sglang/srt/layers/quantization/compressed_tensors/schemes/compressed_tensors_w4a4_mxint4_moe.py b/python/sglang/srt/layers/quantization/compressed_tensors/schemes/compressed_tensors_w4a4_mxint4_moe.py index 47db28201..8419efcb8 100644 --- a/python/sglang/srt/layers/quantization/compressed_tensors/schemes/compressed_tensors_w4a4_mxint4_moe.py +++ b/python/sglang/srt/layers/quantization/compressed_tensors/schemes/compressed_tensors_w4a4_mxint4_moe.py @@ -89,9 +89,8 @@ class CompressedTensorsMxInt4MoE(CompressedTensorsMoEScheme): **extra_weight_attrs, ): assert ( - params_dtype == torch.bfloat16, - f"Params dtype should be torch.bfloat16, but got: {params_dtype}", - ) + params_dtype == torch.bfloat16 + ), f"Params dtype should be torch.bfloat16, but got: {params_dtype}" extra_weight_attrs.update({"quant_method": self.strategy}) w13_weight = torch.nn.Parameter(