GoogleNet does not work on imagenet #559
Comments
sakaia
added a commit
to sakaia/examples
that referenced
this issue
May 13, 2019
GoogLeNet model outputs 3 arguments. This fix takes first tensor (logits). Pull Request resuolved: pytorch#559
Closed
|
We stopped supporting googlenet since it's behavior is funky |
|
Hi msaroufim - sorry to revive an ancient thread. Is googlenet not supported only on ROCm, or not supported on CUDA as well? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
GoogLeNet does not run on imagenet training example.
Since googlenet model have multiple (logits etc) outputs.
To reproduce, do following instruction on Google Colab
'''
!wget https://s3.amazonaws.com/fast-ai-imageclas/imagenette-320.tgz
!tar xzf imagenette-320.tgz`
!git clone http://github.com/pytorch/examples
!pip uninstall -y torchvision
!git clone http://github.com/pytorch/vision
!cd vision;python setup.py install
!python examples/imagenet/main.py -a googlenet -j 2 -b 64 --epochs 10 imagenette-320/
'''
See following discussion
https://discuss.pytorch.org/t/question-about-googlenet/44896
The text was updated successfully, but these errors were encountered: