Missing .pt file

[tnoya001]

Hi. If I follow all your cells and instructions in order, I get this once I reach the second cell of "Generate Single Images"

Using Alias-Free GAN version: 1.0.0
Invalid path /content/drive/MyDrive/colab-gpu-alias-free/alias-free-gan/results/training-000005/000145-epoch-checkpoint.pt is not a .pt model file.

I cannot find that file anywhere or any other .pt file for that matter.

Thanks.

[duskvirkus]

Oh that was just a checkpoint file I was using for testing. I didn't bothering sharing it as a sample because it's was only trained for a little while and it's the not a great model. Happy to share it if you like but you can also download rosinality's ffhq model here: https://aliasfreegan.sfo3.cdn.digitaloceanspaces.com/rosinality-ffhq-800k.pt

I'm still training a 1024 on wikiart landscapes so which I'll put up as the example model as soon as it's done.

Thanks for the feedback I'll try to put up a notebook where the default model works.

[tnoya001]

Completely missed the rosinality pt file. I have it now. All good. Thanks and sorry for the trouble.

[afiaka87]

Oh that was just a checkpoint file I was using for testing. I didn't bothering sharing it as a sample because it's was only trained for a little while and it's the not a great model. Happy to share it if you like but you can also download rosinality's ffhq model here: https://aliasfreegan.sfo3.cdn.digitaloceanspaces.com/rosinality-ffhq-800k.pt

I'm still training a 1024 on wikiart landscapes so which I'll put up as the example model as soon as it's done.

Thanks for the feedback I'll try to put up a notebook where the default model works.

Do you have a timespan on how long the wikiart checkpoint will take to train? That sounds very interesting!

[duskvirkus]

@tnoya001 Latest release has working notebook with rosinality model. https://github.com/duskvirkus/alias-free-gan/blob/stable/notebooks/GPU_Inference_Alias_Free_GAN.ipynb

@afiaka87 Who know's I'm guessing another week or two but I'm having problems with occasional mode collapse and it also depends on what gpus I get on colab for training.