transformers

be2c0e7b - Fixing _pre_quantization_dtype when torch_dtype is None (#36930)