Error when downloading pretrained resnet200

[lwoiceshyn]

Downloading: "https://www.dropbox.com/s/tchq8fbdd4wabjx/fbresnet_200-37304a01b.pth?dl=1" to

/home/leow/.torch/models/fbresnet_200-37304a01b.pth
Traceback (most recent call last):
File "/home/leow/anaconda3/lib/python3.6/urllib/request.py", line 1318, in do_open
encode_chunked=req.has_header('Transfer-encoding'))
File "/home/leow/anaconda3/lib/python3.6/http/client.py", line 1239, in request
self._send_request(method, url, body, headers, encode_chunked)
File "/home/leow/anaconda3/lib/python3.6/http/client.py", line 1285, in _send_request
self.endheaders(body, encode_chunked=encode_chunked)
File "/home/leow/anaconda3/lib/python3.6/http/client.py", line 1234, in endheaders
self._send_output(message_body, encode_chunked=encode_chunked)
File "/home/leow/anaconda3/lib/python3.6/http/client.py", line 1026, in _send_output
self.send(msg)
File "/home/leow/anaconda3/lib/python3.6/http/client.py", line 964, in send
self.connect()
File "/home/leow/anaconda3/lib/python3.6/http/client.py", line 1400, in connect
server_hostname=server_hostname)
File "/home/leow/anaconda3/lib/python3.6/ssl.py", line 401, in wrap_socket
_context=self, _session=session)
File "/home/leow/anaconda3/lib/python3.6/ssl.py", line 808, in init
self.do_handshake()
File "/home/leow/anaconda3/lib/python3.6/ssl.py", line 1061, in do_handshake
self._sslobj.do_handshake()
File "/home/leow/anaconda3/lib/python3.6/ssl.py", line 683, in do_handshake
self._sslobj.do_handshake()
ConnectionResetError: [Errno 104] Connection reset by peer

It seems that the download link doesn't work.

[rwightman]

That's a link to my own dropbox. It worked last time I tried, and still seems to work for me. Try manually if not. There are limits on how much can be downloaded so maybe they were hit. If anything I'll be stripping those links, not replacing them as I do not have a better or alternate hosting solution...

[rwightman]

You can also re-create the pth if you use the https://github.com/clcarwin/convert_torch_to_pytorch converter on the ResNet-200 t7 from https://github.com/facebook/fb.resnet.torch/tree/master/pretrained

[lwoiceshyn]

When I run the torch converter on that file, I get an error because the t7 file has Cuda tensors in it: AttributeError: type object 'torch.cuda.FloatStorage' has no attribute 'from_buffer' Did you run into this issue? Regards, Leo

…

On Tue, Aug 29, 2017 at 4:18 PM, Ross Wightman ***@***.***> wrote: You can also re-create the pth if you use the https://github.com/clcarwin/ convert_torch_to_pytorch converter on the ResNet-200 t7 from https://github.com/facebook/fb.resnet.torch/tree/master/pretrained — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ALkeSyPH_Q0WCa7kkZm_eRIGHePPMNQGks5sdIA9gaJpZM4PGXIQ> .

[rwightman]

Right, that's a problem.

If you've got torch7 installed, from the th console you can do the following to convert it from a cuda model to a cpu float which will work with the conversion script...

th> import 'cudnn'
th> model = torch.load('./resnet-200.t7')
th> model = model:float()
th> torch.save('./resnet-200-cpu.t7', model)