int8量化后，deploy_backend使用tensorrt，转onnx后转tensorrt，有很多Missing scale and zero-point for tensor是为什么

[mandylyin]

问题描述
int8量化后，deploy_backend使用tensorrt，转onnx后转tensorrt，有很多Missing scale and zero-point for tensor

更多信息 :

• 用于部署的推理引擎: tensorrt8.6
• Paddle2ONNX版本: 1.3.1
• paddle版本：2.6
• PaddleSeg：2.9
• 量化后的模型文件：
  链接：https://pan.baidu.com/s/1qiNlDMr0Haq3RUWTFhqe9g?pwd=585d
  提取码：585d
  模型是PaddleSeg的rtformer，QAT量化。
  paddel2onnx命令：
  paddle2onnx --model_dir=./output_quant_for_trt/ --model_filename=model.pdmodel --params=model.pdiparams --save_file=./output_quant_for_trt/model.onnx --opset_version=16 --enable_onnx_checker=True --deploy_backend tensorrt --save_calibration_file ./output_quant_for_trt/calibration.cache

错误信息：
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_0.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_0.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_1.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_1.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_2.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_3.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.5, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_4.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_5.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.7, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_6.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_7.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_8.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.9, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_9.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_10.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.11, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_11.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_12.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_13.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.13, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_14.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_15.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.15, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_16.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor bilinear_interp_v2_0.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.17, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_17.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_18.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.19, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor relu_16.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor relu_17.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor relu_18.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_20.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_20.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_2.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 75) [Softmax]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor softmax_0.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor ReduceSum.1, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Div.1, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_3.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_21.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.21, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_22.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 84) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 85) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Div.3, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Erf.1, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 88) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 89) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.25, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Mul.1, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 92) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 93) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_23.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.29, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_22.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor bilinear_interp_v2_1.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.31, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_23.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor split_0.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor split_0.tmp_1, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor transpose_0.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_6.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_7.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_24.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_8.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_26.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_9.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 118) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 119) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Mul.4, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 121) [Softmax]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor softmax_1.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_10.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_27.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_11.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.33, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_28.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 129) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 130) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Div.5, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Erf.3, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 133) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 134) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.37, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Mul.6, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 137) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 138) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_29.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.41, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_26.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_27.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.43, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor relu_22.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor relu_23.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_28.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_32.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_14.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 154) [Softmax]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor softmax_2.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor ReduceSum.3, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Div.7, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_15.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_33.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.45, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_34.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 163) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 164) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Div.9, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Erf.5, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 167) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 168) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.49, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Mul.9, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 171) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 172) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_35.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.53, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_30.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor bilinear_interp_v2_2.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.55, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_31.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor split_1.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor split_1.tmp_1, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor transpose_1.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_18.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_19.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_32.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_20.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_38.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_21.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 197) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 198) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Mul.12, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 200) [Softmax]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor softmax_3.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_22.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_39.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor reshape2_23.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.57, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_40.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 208) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 209) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Div.11, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Erf.7, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 212) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 213) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.61, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Mul.14, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 216) [Constant]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 217) [Shuffle]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_41.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.65, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_34.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_35.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.67, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_36.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor pool2d_0.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_37.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor bilinear_interp_v2_3.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.69, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_38.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor pool2d_1.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_39.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor bilinear_interp_v2_4.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.71, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_40.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor pool2d_2.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_41.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor bilinear_interp_v2_5.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.73, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_42.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor pool2d_3.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_43.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor bilinear_interp_v2_6.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.75, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_44.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Concat.8, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_45.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_46.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Add.77, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor bilinear_interp_v2_7.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor Concat.11, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_47.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor batch_norm_48.tmp_2, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor conv2d_56.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor bilinear_interp_v2_8.tmp_0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 308) [TopK]_output, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor (Unnamed Layer* 308) [TopK]_output_1, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor ArgMax.1, expect fall back to non-int8 implementation for any layer consuming or producing given tensor
[01/03/2025-05:34:56] [TRT] [W] Missing scale and zero-point for tensor save_infer_model/scale_0.tmp_0.0, expect fall back to non-int8 implementation for any layer consuming or producing given tensor

导出的calibration.cache:
TRT-8XXX-EntropyCalibration2
x.quantized.dequantized: 40008102
x: 40008102
relu_9.tmp_0.quantized.dequantized.0: 3e06f7a3
relu_9.tmp_0.quantized.dequantized: 3e06f7a3
relu_9.tmp_0: 3e06f7a3
relu_8.tmp_0.quantized.dequantized: 3d3cd8d8
relu_8.tmp_0: 3d3cd8d8
relu_7.tmp_0.quantized.dequantized: 3db6ed15
relu_7.tmp_0: 3db6ed15
relu_6.tmp_0.quantized.dequantized: 3d6f53ad
relu_6.tmp_0: 3d6f53ad
relu_5.tmp_0.quantized.dequantized.0: 3e3d5019
relu_5.tmp_0.quantized.dequantized: 3e3d5019
relu_5.tmp_0: 3e3d5019
relu_4.tmp_0.quantized.dequantized: 3dbe8e70
relu_4.tmp_0: 3dbe8e70
relu_39.tmp_0.quantized.dequantized: 3da30a54
relu_39.tmp_0: 3da30a54
relu_38.tmp_0.quantized.dequantized: 3e24fd50
relu_38.tmp_0: 3e24fd50
relu_37.tmp_0.quantized.dequantized: 3dba6e2e
relu_37.tmp_0: 3dba6e2e
relu_36.tmp_0.quantized.dequantized: 3d6c3bb1
relu_36.tmp_0: 3d6c3bb1
relu_35.tmp_0.quantized.dequantized: 3d8375fb
relu_35.tmp_0: 3d8375fb
relu_34.tmp_0.quantized.dequantized: 3cee0f49
relu_34.tmp_0: 3cee0f49
relu_33.tmp_0.quantized.dequantized: 3d75091e
relu_33.tmp_0: 3d75091e
relu_32.tmp_0.quantized.dequantized: 3ce96e64
relu_32.tmp_0: 3ce96e64
relu_31.tmp_0.quantized.dequantized: 3da0d844
relu_31.tmp_0: 3da0d844
relu_30.tmp_0.quantized.dequantized: 3d5d1e2f
relu_30.tmp_0: 3d5d1e2f
relu_3.tmp_0.quantized.dequantized: 3e2e2a50
relu_3.tmp_0: 3e2e2a50
relu_29.tmp_0.quantized.dequantized: 3d8e8cef
relu_29.tmp_0: 3d8e8cef
relu_28.tmp_0.quantized.dequantized: 3d34d6e5
relu_28.tmp_0: 3d34d6e5
relu_27.tmp_0.quantized.dequantized: 3da917c3
relu_27.tmp_0: 3da917c3
relu_26.tmp_0.quantized.dequantized: 3e055356
relu_26.tmp_0: 3e055356
relu_25.tmp_0.quantized.dequantized: 3dd3d1c5
relu_25.tmp_0: 3dd3d1c5
relu_24.tmp_0.quantized.dequantized: 3de9ebfa
relu_24.tmp_0: 3de9ebfa
relu_21.tmp_0.quantized.dequantized: 3e31cd00
relu_21.tmp_0: 3e31cd00
relu_20.tmp_0.quantized.dequantized: 3e1bdd95
relu_20.tmp_0: 3e1bdd95
relu_2.tmp_0.quantized.dequantized: 3e04a01b
relu_2.tmp_0: 3e04a01b
relu_19.tmp_0.quantized.dequantized: 3cf084b5
relu_19.tmp_0: 3cf084b5
relu_15.tmp_0.quantized.dequantized: 3d54e7e6
relu_15.tmp_0: 3d54e7e6
relu_14.tmp_0.quantized.dequantized: 3e2c78a6
relu_14.tmp_0: 3e2c78a6
relu_13.tmp_0.quantized.dequantized: 3dac3acc
relu_13.tmp_0: 3dac3acc
relu_12.tmp_0.quantized.dequantized: 3d23f37b
relu_12.tmp_0: 3d23f37b
relu_11.tmp_0.quantized.dequantized: 3d6ff790
relu_11.tmp_0: 3d6ff790
relu_10.tmp_0.quantized.dequantized: 3d3fce83
relu_10.tmp_0: 3d3fce83
relu_1.tmp_0.quantized.dequantized: 3e26c865
relu_1.tmp_0: 3e26c865
relu_0.tmp_0.quantized.dequantized: 3e42f76a
relu_0.tmp_0: 3e42f76a
max_pool2d_with_index_1.tmp_0.quantized.dequantized: 3da0fb91
max_pool2d_with_index_1.tmp_0: 3da0fb91
max_pool2d_with_index_0.tmp_0.quantized.dequantized: 3cfc7d2f
max_pool2d_with_index_0.tmp_0: 3cfc7d2f
gelu_3.tmp_0.quantized.dequantized: 3df07eff
gelu_3.tmp_0: 3df07eff
gelu_2.tmp_0.quantized.dequantized: 3eab25ef
gelu_2.tmp_0: 3eab25ef
gelu_1.tmp_0.quantized.dequantized: 3ddf3ba1
gelu_1.tmp_0: 3ddf3ba1
gelu_0.tmp_0.quantized.dequantized: 3ec5ba52
gelu_0.tmp_0: 3ec5ba52
conv2d_9.tmp_0.tmp: 3ef6b973
conv2d_9.tmp_0: 3ef6b973
conv2d_8.tmp_0.tmp: 3e4498ce
conv2d_8.tmp_0: 3e4498ce
conv2d_7.tmp_0.tmp: 3e605dde
conv2d_7.tmp_0: 3e605dde
conv2d_6.tmp_0.tmp: 3f1dc67e
conv2d_6.tmp_0: 3f1dc67e
conv2d_56.tmp_1.tmp: 3eabf0a5
conv2d_56.tmp_1: 3eabf0a5
conv2d_55.tmp_0.tmp: 4007a1a0
conv2d_55.tmp_0: 4007a1a0
conv2d_54.tmp_0.tmp: 3e91fac6
conv2d_54.tmp_0: 3e91fac6
conv2d_53.tmp_0.tmp: 3e95402c
conv2d_53.tmp_0: 3e95402c
conv2d_52.tmp_0.tmp: 3f323ef3
conv2d_52.tmp_0: 3f323ef3
conv2d_51.tmp_0.tmp: 3ddfb51d
conv2d_51.tmp_0: 3ddfb51d
conv2d_50.tmp_0.tmp: 3f0989c8
conv2d_50.tmp_0: 3f0989c8
conv2d_5.tmp_0.tmp: 3e688fa3
conv2d_5.tmp_0: 3e688fa3
conv2d_49.tmp_0.tmp: 3da07903
conv2d_49.tmp_0: 3da07903
conv2d_48.tmp_0.tmp: 3f1bf9a8
conv2d_48.tmp_0: 3f1bf9a8
conv2d_47.tmp_0.tmp: 3e6aa4fc
conv2d_47.tmp_0: 3e6aa4fc
conv2d_46.tmp_0.tmp: 3f714f69
conv2d_46.tmp_0: 3f714f69
conv2d_45.tmp_0.tmp: 3e1485f2
conv2d_45.tmp_0: 3e1485f2
conv2d_44.tmp_0.tmp: 3e8f0dfa
conv2d_44.tmp_0: 3e8f0dfa
conv2d_43.tmp_0.tmp: 40427270
conv2d_43.tmp_0: 40427270
conv2d_42.tmp_0.tmp: 3f31707a
conv2d_42.tmp_0: 3f31707a
conv2d_41.tmp_1.tmp: 3e5baa22
conv2d_41.tmp_1: 3e5baa22
conv2d_40.tmp_1.tmp: 3eb15a38
conv2d_40.tmp_1: 3eb15a38
conv2d_4.tmp_0.tmp: 3ef83403
conv2d_4.tmp_0: 3ef83403
conv2d_37.tmp_0.tmp: 3e42f995
conv2d_37.tmp_0: 3e42f995
conv2d_36.tmp_0.tmp: 3e885d32
conv2d_36.tmp_0: 3e885d32
conv2d_35.tmp_1.tmp: 3f82e625
conv2d_35.tmp_1: 3f82e625
conv2d_34.tmp_1.tmp: 3f415c20
conv2d_34.tmp_1: 3f415c20
conv2d_31.tmp_0.tmp: 402dc4c6
conv2d_31.tmp_0: 402dc4c6
conv2d_30.tmp_0.tmp: 3f39c246
conv2d_30.tmp_0: 3f39c246
conv2d_3.tmp_0.tmp: 3ec4cf30
conv2d_3.tmp_0: 3ec4cf30
conv2d_29.tmp_1.tmp: 3e4b392b
conv2d_29.tmp_1: 3e4b392b
conv2d_28.tmp_1.tmp: 3e8991fa
conv2d_28.tmp_1: 3e8991fa
conv2d_25.tmp_0.tmp: 3e4c862c
conv2d_25.tmp_0: 3e4c862c
conv2d_24.tmp_0.tmp: 3d09cdf2
conv2d_24.tmp_0: 3d09cdf2
conv2d_23.tmp_1.tmp: 403cb514
conv2d_23.tmp_1: 403cb514
conv2d_22.tmp_1.tmp: 3f11b66d
conv2d_22.tmp_1: 3f11b66d
conv2d_2.tmp_0.tmp: 3edad7af
conv2d_2.tmp_0: 3edad7af
conv2d_19.tmp_0.tmp: 3e67af05
conv2d_19.tmp_0: 3e67af05
conv2d_18.tmp_0.tmp: 3e4dd594
conv2d_18.tmp_0: 3e4dd594
conv2d_17.tmp_0.tmp: 4044946a
conv2d_17.tmp_0: 4044946a
conv2d_16.tmp_0.tmp: 3e2543c3
conv2d_16.tmp_0: 3e2543c3
conv2d_15.tmp_0.tmp: 3e5a8758
conv2d_15.tmp_0: 3e5a8758
conv2d_14.tmp_0.tmp: 3eba2e5e
conv2d_14.tmp_0: 3eba2e5e
conv2d_13.tmp_0.tmp: 3e19d493
conv2d_13.tmp_0: 3e19d493
conv2d_12.tmp_0.tmp: 3e4b4de5
conv2d_12.tmp_0: 3e4b4de5
conv2d_11.tmp_0.tmp: 3f256097
conv2d_11.tmp_0: 3f256097
conv2d_10.tmp_0.tmp: 3e4fc34e
conv2d_10.tmp_0: 3e4fc34e
conv2d_1.tmp_1.tmp: 3ef71da4
conv2d_1.tmp_1: 3ef71da4
conv2d_0.tmp_1.tmp: 4102bafe
conv2d_0.tmp_1: 4102bafe
batch_norm_33.tmp_2.quantized.dequantized: 3da6047f
batch_norm_33.tmp_2: 3da6047f
batch_norm_29.tmp_2.quantized.dequantized: 3e2ff57b
batch_norm_29.tmp_2: 3e2ff57b
batch_norm_25.tmp_2.quantized.dequantized: 3d98a698
batch_norm_25.tmp_2: 3d98a698
batch_norm_21.tmp_2.quantized.dequantized: 3d1b5e66
batch_norm_21.tmp_2: 3d1b5e66
batch_norm_19.tmp_2.quantized.dequantized: 3e1d2261
batch_norm_19.tmp_2: 3e1d2261
auto.cast.9: 3f0989c8
auto.cast.8: 3f1bf9a8
auto.cast.7: 3f714f69
auto.cast.6: 3e8f0dfa
auto.cast.5: 3eb15a38
auto.cast.3: 3f415c20
auto.cast.2: 3e8991fa
auto.cast.10: 3f323ef3
auto.cast.0: 3f11b66d
Add.79: 3eabf0a5
Add.63: 3e5baa22
Add.59: 3eb15a38
Add.51: 3f82e625
Add.47: 3f415c20
Add.39: 3e4b392b
Add.35: 3e8991fa
Add.3: 3ef71da4
Add.27: 403cb514
Add.23: 3f11b66d
Add.1: 4102bafe

[Jiang-Jia-Jun]

应该是量化格式TensorRT识别存在问题。 我看上面的转换只是一个WARNING信息，你在转换完后，测试下推理精度，如果没问题则可使用。 如果有问题，建议不要将量化后的模型转ONNX，而是使用Float32模型转ONNX，再转TRT