UnpicklingError: invalid load key, '<'.

[josejhlee]

Hello authors,

I have an error when using the pretrained checkpoints when torch.load(args.ckpt) is run in generate.py (or any piece of code with the function). I tried re-downloading the model as proposed in this issue. The only change I made to accommodate my CUDA 11 version was to instal pytorch with conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch. Thanks for you attention, it is much appreciated.

[josejhlee]

Solved by manually downloading through the google drive link!

[brando90]

related: https://github.com/learnables/learn2learn/issues/310

[chLFF]

Solved by manually downloading through the google drive link!

Can you plz share the google drive link or the checkpoints file? The links in download.sh are disabled. And i can not find the google drive link.

[josejhlee]

@chLFF I am able to open the links without a problem. Which want do you want? And do the authors consent in this?

[chLFF]

@josejhlee The link says 'i'm sorry. the link has expired.' I want celeba_hq_256_8x8.pt and celeba_hq_5M_256_8x8.pt, thank you. I didn't get consent yet but i think this work is open-source.

[josejhlee]

@chLFF Yeah, doesn't seem like they closed on purpose. Let me know if these are the ones that you asked for. celeba_hq_8x8_20M_revised https://drive.google.com/file/d/1BCvVb8ipydQMFU4ZsuMcW3pKSubtALd3/view?usp=sharing celeba_hq_8x8_5M_revised: https://drive.google.com/file/d/17OxdwVK--WjZXrHe1iupMQSP9VscaHm0/view?usp=sharing

[chLFF]

@josejhlee 'celeba_hq_8x8_20M_revised' is one of i need, thanks! Do you have 5M-image-trained model of CelebA-256? I believe its name is celeba_hq_5M_256_8x8.pt.

[josejhlee]

@chLFF My bad, I edited the link above!

[chLFF]

@josejhlee This helps me a lot, thank you! 고맙다!