[BUG] Issue title...Pretrained model URL is invalid, using random initialization.

[BCWang93]

Describe the bug
Hi, when I load the model 'resnet200d',I get this error.This mean the pretrained model weight is not use?Thanks!

To Reproduce
Steps to reproduce the behavior:
1.
2.

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

• OS: [e.g. Windows 10, Ubuntu 18.04]
• This repository version [e.g. pip 0.3.1 or commit ref]
• PyTorch version w/ CUDA/cuDNN [e.g. from conda list, 1.7.0 py3.8_cuda11.0.221_cudnn8.0.3_0]

Additional context
Add any other context about the problem here.

[rwightman]

The weights are there, you may be using an old version or the pypi version, pypi (pip) hasn't been updated yet.

[BCWang93]

The weights are there, you may be using an old version or the pypi version, pypi (pip) hasn't been updated yet.

I use the timm version is 0.3.2,but I only can't get the weight!And I find,the efficientnet_b5 is also get this problem!

[linhduongtuan]

Ross has not trained his own efficientnet_b5. He ports weights (B4-L2, ap, and ns) from the original TensorFlow source so you should use tf_efficientnet_b5.

…

On 28 Dec 2020, at 11:24, Bcw93 ***@***.***> wrote: The weights are there, you may be using an old version or the pypi version, pypi (pip) hasn't been updated yet. I use the timm version is 0.3.2,but I only can't get the weight!And I find,the efficientnet_b5 is also get this problem! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#316 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFKZ27FIYKD2VBP365QMNXDSXACBVANCNFSM4VCZSZGA>.

[rwightman]

@BCWang93
There are no efficientnet_b5 weights, there are also no resnet200 weights in case you missed the d

timm.list_models('*efficientnet_b5*', pretrained=True)
['tf_efficientnet_b5', 'tf_efficientnet_b5_ap', 'tf_efficientnet_b5_ns']

timm.list_models('resnet*d', pretrained=True)
['resnet18d', 'resnet26d', 'resnet34d', 'resnet50d', 'resnet101d', 'resnet152d', 'resnet200d']

timm.models.resnet.default_cfgs['resnet200d']['url']
'https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet200d_ra2-bdba9bf9.pth'

If any of those commands in a console (with timm imported obviously) do not yield the same output you have the wrong version of timm in your environment .... as I've mentioned I tested it, that's how I got my validation result. If the above is the same and it still doesn't work I really have no clue what's going on, you'll have to debug further and let me know. Or you can download the weight from the url with wget/curl and load checkpoint manually.

[BCWang93]

@BCWang93
There are no efficientnet_b5 weights, there are also no resnet200 weights in case you missed the d

timm.list_models('*efficientnet_b5*', pretrained=True)
['tf_efficientnet_b5', 'tf_efficientnet_b5_ap', 'tf_efficientnet_b5_ns']

timm.list_models('resnet*d', pretrained=True)
['resnet18d', 'resnet26d', 'resnet34d', 'resnet50d', 'resnet101d', 'resnet152d', 'resnet200d']

timm.models.resnet.default_cfgs['resnet200d']['url']
'https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet200d_ra2-bdba9bf9.pth'

If any of those commands in a console (with timm imported obviously) do not yield the same output you have the wrong version of timm in your environment .... as I've mentioned I tested it, that's how I got my validation result. If the above is the same and it still doesn't work I really have no clue what's going on, you'll have to debug further and let me know. Or you can download the weight from the url with wget/curl and load checkpoint manually.

Hi,I also can't get the url.This is my output.

[BCWang93]

@BCWang93
There are no efficientnet_b5 weights, there are also no resnet200 weights in case you missed the d

timm.list_models('*efficientnet_b5*', pretrained=True)
['tf_efficientnet_b5', 'tf_efficientnet_b5_ap', 'tf_efficientnet_b5_ns']

timm.list_models('resnet*d', pretrained=True)
['resnet18d', 'resnet26d', 'resnet34d', 'resnet50d', 'resnet101d', 'resnet152d', 'resnet200d']

timm.models.resnet.default_cfgs['resnet200d']['url']
'https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet200d_ra2-bdba9bf9.pth'

If any of those commands in a console (with timm imported obviously) do not yield the same output you have the wrong version of timm in your environment .... as I've mentioned I tested it, that's how I got my validation result. If the above is the same and it still doesn't work I really have no clue what's going on, you'll have to debug further and let me know. Or you can download the weight from the url with wget/curl and load checkpoint manually.

Another problem,I had download the pretrained weight of 'resnet200d' manually,but when I create model,it is still give me this warnings 'Pretrained model URL is invalid, using random initialization.'.So,I don't know what happend!

[BCWang93]

@BCWang93
There are no efficientnet_b5 weights, there are also no resnet200 weights in case you missed the d

timm.list_models('*efficientnet_b5*', pretrained=True)
['tf_efficientnet_b5', 'tf_efficientnet_b5_ap', 'tf_efficientnet_b5_ns']

timm.list_models('resnet*d', pretrained=True)
['resnet18d', 'resnet26d', 'resnet34d', 'resnet50d', 'resnet101d', 'resnet152d', 'resnet200d']

timm.models.resnet.default_cfgs['resnet200d']['url']
'https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet200d_ra2-bdba9bf9.pth'

If any of those commands in a console (with timm imported obviously) do not yield the same output you have the wrong version of timm in your environment .... as I've mentioned I tested it, that's how I got my validation result. If the above is the same and it still doesn't work I really have no clue what's going on, you'll have to debug further and let me know. Or you can download the weight from the url with wget/curl and load checkpoint manually.

Otherwise,if I am not use 'pip', I have other methods to install 'timm' package?Thanks!

[linhduongtuan]

Let try pip install git+https://github.com/rwightman/pytorch-image-models.git <git+https://github.com/rwightman/pytorch-image-models.git> It is usually more updated than that of pip install timm. Ross occationally updates pip timm. Good luck. Linh

…

On 30 Dec 2020, at 10:32, Bcw93 ***@***.***> wrote: @BCWang93 <https://github.com/BCWang93> There are no efficientnet_b5 weights, there are also no resnet200 weights in case you missed the d timm.list_models('*efficientnet_b5*', pretrained=True) ['tf_efficientnet_b5', 'tf_efficientnet_b5_ap', 'tf_efficientnet_b5_ns'] timm.list_models('resnet*d', pretrained=True) ['resnet18d', 'resnet26d', 'resnet34d', 'resnet50d', 'resnet101d', 'resnet152d', 'resnet200d'] timm.models.resnet.default_cfgs['resnet200d']['url'] 'https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet200d_ra2-bdba9bf9.pth' If any of those commands in a console (with timm imported obviously) do not yield the same output you have the wrong version of timm in your environment .... as I've mentioned I tested it, that's how I got my validation result. If the above is the same and it still doesn't work I really have no clue what's going on, you'll have to debug further and let me know. Or you can download the weight from the url with wget/curl and load checkpoint manually. Otherwise,if I am not use 'pip', I have other methods to install 'timm' package?Thanks! — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#316 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFKZ27G2RI6EILVM2PYGW4TSXKNOFANCNFSM4VCZSZGA>.

[BCWang93]

Let try pip install git+https://github.com/rwightman/pytorch-image-models.git git+https://github.com/rwightman/pytorch-image-models.git It is usually more updated than that of pip install timm. Ross occationally updates pip timm. Good luck. Linh
…
On 30 Dec 2020, at 10:32, Bcw93 @.**> wrote: @BCWang93 https://github.com/BCWang93 There are no efficientnet_b5 weights, there are also no resnet200 weights in case you missed the d timm.list_models('efficientnet_b5', pretrained=True) ['tf_efficientnet_b5', 'tf_efficientnet_b5_ap', 'tf_efficientnet_b5_ns'] timm.list_models('resnetd', pretrained=True) ['resnet18d', 'resnet26d', 'resnet34d', 'resnet50d', 'resnet101d', 'resnet152d', 'resnet200d'] timm.models.resnet.default_cfgs['resnet200d']['url'] 'https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet200d_ra2-bdba9bf9.pth' If any of those commands in a console (with timm imported obviously) do not yield the same output you have the wrong version of timm in your environment .... as I've mentioned I tested it, that's how I got my validation result. If the above is the same and it still doesn't work I really have no clue what's going on, you'll have to debug further and let me know. Or you can download the weight from the url with wget/curl and load checkpoint manually. Otherwise,if I am not use 'pip', I have other methods to install 'timm' package?Thanks! — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#316 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFKZ27G2RI6EILVM2PYGW4TSXKNOFANCNFSM4VCZSZGA.

Thanks!I will try it!

[BCWang93]

Let try pip install git+https://github.com/rwightman/pytorch-image-models.git git+https://github.com/rwightman/pytorch-image-models.git It is usually more updated than that of pip install timm. Ross occationally updates pip timm. Good luck. Linh
…
On 30 Dec 2020, at 10:32, Bcw93 @.**> wrote: @BCWang93 https://github.com/BCWang93 There are no efficientnet_b5 weights, there are also no resnet200 weights in case you missed the d timm.list_models('efficientnet_b5', pretrained=True) ['tf_efficientnet_b5', 'tf_efficientnet_b5_ap', 'tf_efficientnet_b5_ns'] timm.list_models('resnetd', pretrained=True) ['resnet18d', 'resnet26d', 'resnet34d', 'resnet50d', 'resnet101d', 'resnet152d', 'resnet200d'] timm.models.resnet.default_cfgs['resnet200d']['url'] 'https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/resnet200d_ra2-bdba9bf9.pth' If any of those commands in a console (with timm imported obviously) do not yield the same output you have the wrong version of timm in your environment .... as I've mentioned I tested it, that's how I got my validation result. If the above is the same and it still doesn't work I really have no clue what's going on, you'll have to debug further and let me know. Or you can download the weight from the url with wget/curl and load checkpoint manually. Otherwise,if I am not use 'pip', I have other methods to install 'timm' package?Thanks! — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#316 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFKZ27G2RI6EILVM2PYGW4TSXKNOFANCNFSM4VCZSZGA.

Hi,can I install this package use the 'python install setup.py'?Thanks!