cannot download, how to change the path?

[Brandon Ng]

Hello, I would like to ask if using this code

from resnest.torch import resnest50
net = resnest50(pretrained=True)

the code will download of the pre-trained model stored on wasabi,
Now I have downloaded the pre-trained model file from elsewhere, how do I change the path from this code?
Because this code will read the pre-trained model from /root/.cache/torch/hub/checkpoints by default

net = resnest50(pretrained=True)

[Hang Zhang]

Please reinstall the newest package on github. The links are already updated.

[Brandon Ng]

Thank you so much for your reply.

I ran the command

!pip install resnest --pre

But the error message shows that the old download link is still being used

# test HTTP Status
from resnest.torch import resnest50
net = resnest50(pretrained=True)

Downloading: "https://s3.us-west-1.wasabisys.com/resnest/torch/resnest50-528c19ca.pth" to /root/.cache/torch/hub/checkpoints/resnest50-528c19ca.pth
---------------------------------------------------------------------------
HTTPError                                 Traceback (most recent call last)
<ipython-input-15-06d7e8179e05> in <module>
      1 # test HTTP Status
      2 from resnest.torch import resnest50
----> 3 net = resnest50(pretrained=True)

[Hang Zhang]

Could you do
pip install git+https://github.com/zhanghang1989/ResNeSt?

[Brandon Ng]

I've done what you said

pip install git+https://github.com/zhanghang1989/ResNeSt

My question is mainly about....Is It use this code to import the pre-trained model?

from resnest.torch import resnest50
net = resnest50(pretrained=True)

[Hang Zhang]

Yes, that's correct.

You can also use torch hub:

import torch
# get list of models
torch.hub.list('zhanghang1989/ResNeSt', force_reload=True)

# load pretrained models, using ResNeSt-50 as an example
net = torch.hub.load('zhanghang1989/ResNeSt', 'resnest50', pretrained=True)

The models can also be download manually from:

https://github.com/zhanghang1989/ResNeSt/releases/tag/weights_step1

[Brandon Ng]

Thank you so much~