suncheng-s

Please. **Describe the bug** I tried to export the TimeSformer onnx using pytorch2onnx.py. python3 tools/deployment/pytorch2onnx.py work_dirs/timesformer_divST_16-2-1_violenceSet_613/timesformer_divST_8x32x1_15e_violenceSet_rgb.py work_dirs/timesformer_divST_16-2-1_violenceSet_613/best_top1_acc_epoch_5.pth --shape 1 3 16 224 224 --verify (mmlab) root@b434ab344879:/home/cheng/workspace/actionClassification/mmaction2# python3 tools/deployment/pytorch2onnx.py work_dirs/timesformer_divST_16-2-1_violenceSet_613/timesformer_divST_8x32x1_15e_violenceSet_rgb.py work_dirs/timesformer_divST_16-2-1_violenceSet_613/best_top1_acc_epoch_5.pth...

转化ONNX时如何得到与推理相同的输出格式？

10

转化ONNX时如何得到与推理相同的输出格式？目前转化为ONNX时，输出为三个特征图的结果。怎么修改能得到和推理时一样的输出结果呢？如下图所示。  我修改了model.model[-1].export = False，输出格式相同，但是提示错误：  请问该如何解决？

Could you please provide the link of LSTM model again?

1

Could you please provide the link of LSTM model again?