AttributeError: module 'torch.onnx' has no attribute 'set_training'

[keosu]

pytorch version: 1.6.0

import sys
import torch
import tensorwatch as tw
import torchvision.models

alexnet_model = torchvision.models.alexnet()
tw.draw_model(alexnet_model, [1, 3, 224, 224])

[btdan]

I have the same question, Is there anyone solve it?Thanks very much.

[jian-ke]

Maybe we should mofify some code.

[kuku-miku]

Do we have any solution? Thank you,

[YunhaoDu]

just change the source code is OK

[kuku-miku]

just change the source code is OK

so basically you mean to just change at line 85 in /home/ant/anaconda3/lib/python3.8/site-packages/tensorwatch/model_graph/hiddenlayerhiddenlayer/summary_graph.py i.e.

    #with torch.onnx.set_training(model_clone, False):
    #with torch.onnx.select_model_mode_for_export(model_clone, None):
    with torch.onnx.select_model_mode_for_export(model_clone, False):

        device = distiller.model_device(model_clone)
        dummy_input = distiller.convert_tensors_recursively_to(dummy_input, device=device)
        self.dummy_input = dummy_input

well, above changes give another error as below,
File "/home/ant/anaconda3/lib/python3.8/site-packages/torch/onnx/utils.py", line 1083, in _node_getitem
return getattr(self, sel)(k)
AttributeError: 'torch._C.Node' object has no attribute 'ival'
:(

[YunhaoDu]

just change the source code is OK

so basically you mean to just change at line 85 in /home/ant/anaconda3/lib/python3.8/site-packages/tensorwatch/model_graph/hiddenlayerhiddenlayer/summary_graph.py i.e.

    #with torch.onnx.set_training(model_clone, False):
    #with torch.onnx.select_model_mode_for_export(model_clone, None):
    with torch.onnx.select_model_mode_for_export(model_clone, False):

        device = distiller.model_device(model_clone)
        dummy_input = distiller.convert_tensors_recursively_to(dummy_input, device=device)
        self.dummy_input = dummy_input

well, above changes give another error as below,
File "/home/ant/anaconda3/lib/python3.8/site-packages/torch/onnx/utils.py", line 1083, in _node_getitem
return getattr(self, sel)(k)
AttributeError: 'torch._C.Node' object has no attribute 'ival'
:(

You're right. There are troublesome problems. I've tried PyTorch v0.4, 1.1, 1.6. Now I've given up using tensorwatch.
Instead, PyTorchViz, tensorboardX and netron are also good tools.

[wyuzyf]

Have you solved the problem?

[Bowenwu1]

same problem

[holestine]

I had this issue due to an old version of TensorBoardX, installing version 2.2 fixed it for me.

[ZyangLee]

Same problem

[102757017]

Same problem

[CheapMeow]

I had this issue due to an old version of TensorBoardX, installing version 2.2 fixed it for me.

Seems don't work for me.

just change the source code is OK

I meet error:

Traceback (most recent call last):
  File "D:\Work\PyCharmProject\nlp-tutorial-master\venv\lib\site-packages\tensorwatch\model_graph\hiddenlayer\pytorch_draw_model.py", line 63, in draw_img_classifier
    g = SummaryGraph(non_para_model, dummy_input)
  File "D:\Work\PyCharmProject\nlp-tutorial-master\venv\lib\site-packages\tensorwatch\model_graph\hiddenlayer\summary_graph.py", line 86, in __init__
    with torch.onnx.select_model_mode_for_export(model_clone, False):
  File "C:\Users\CheapMeow\AppData\Local\Programs\Python\Python310\lib\contextlib.py", line 135, in __enter__
    return next(self.gen)
  File "D:\Work\PyCharmProject\nlp-tutorial-master\venv\lib\site-packages\torch\onnx\utils.py", line 88, in select_model_mode_for_export
    raise TypeError(
TypeError: 'mode' should be a torch.onnx.TrainingMode enum, but got '<class 'bool'>'.

[JasonYangCode]

Same problem

Same problem