* Collect output scale for quantized op and fused op * Post_training_quantizaion sets batch_generator to support lod tensor
拖放文件到此处或点击上传
