search

pytorch / ignite

High-level library to help with training and evaluating neural networks in PyTorch flexibly and transparently.
https://pytorch-ignite.ai
BSD 3-Clause "New" or "Revised" License
4.52k stars 615 forks source link

I got an error in EfficientNet notebook example #591

Closed sooperset closed 5 years ago

sooperset commented 5 years ago

torch : '1.2.0+cu92' ignite: '0.2.0' tensorboardX: '1.8'

Hi, I just followed efficientNet notebook example in the git.

x = torch.rand(4, 3, 224, 224)
graph_def = graph(model, x, operator_export_type='RAW')

In this part, I got an error

/opt/conda/lib/python3.6/site-packages/ipykernel_launcher.py:61: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!

AttributeError Traceback (most recent call last)

in 1 x = torch.rand(4, 3, 224, 224) ----> 2 graph_def = graph(model, x, operator_export_type='RAW') /opt/conda/lib/python3.6/site-packages/tensorboardX/pytorch_graph.py in graph(model, args, verbose, **kwargs) 253 if verbose: 254 print(graph)

My tensorboardX version is different or is there anything I can fix?

vfdev-5 commented 5 years ago

@Soo95 thanks for the feedback! Yes, most probably my tbX version was 1.7 at the moment of writing. Anyway, I'll check and fix this error in the coming days.

sooperset commented 5 years ago

@vfdev-5 https://github.com/lanpa/tensorboardX/issues/483 This is the same issue with me. the tensorboard graph has a problem after pytorch 1.2 version. I think I can do nothing unless this bug is fixed from tensorboard.