search

didriknielsen / survae_flows

Code for paper "SurVAE Flows: Surjections to Bridge the Gap between VAEs and Flows"
MIT License
283 stars 34 forks source link

Can you upload pretrained model for imagenet please? #9

Open moonman925 opened 3 years ago

moonman925 commented 3 years ago

Hi, there! I really like your work! For some reason I want to run some experiment upon the model given by your repo in experiments/imagefolder.

However, it's so hard to train such a big model due to limited computational resources, so, if it is possible, can you please upload the pretrained model for imagenet64*64? Thanks!!

didriknielsen commented 3 years ago

Hey! Thanks!

I'd be happy to share the pre-trained models. Do you know of any good sites for sharing large files publicly? Each model is ~1GB.

schuhschuh commented 3 years ago

Could this be an option? https://docs.github.com/en/github/managing-large-files/distributing-large-binaries

moonman925 commented 3 years ago

I think github release and google drive would be very good choices. 😋

didriknielsen commented 3 years ago

Thanks for the suggestions! I will try GitHub release. I'm a bit busy for the moment, but I will release the weights early next week.

didriknielsen commented 3 years ago

Released the weights now! https://github.com/didriknielsen/survae_flows/releases/tag/v1.0.0

Fangwq commented 2 years ago

@didriknielsen, by the way, where to download the imagnet data? It seems that the data can not be found any more.

Downloading http://image-net.org/small/train_32x32.tar
urllib.error.HTTPError: HTTP Error 404: Not Found
didriknielsen commented 2 years ago

That is a good question. I'm not sure what the best source is currently.

It seems available as a torrent here, which might be the best as long as there are seeders: Imagenet 32x32: https://academictorrents.com/details/bf62f5051ef878b9c357e6221e879629a9b4b172 Imagenet 64x64: https://academictorrents.com/details/96816a530ee002254d29bf7a61c0c158d3dedc3b

You can also download them using the description here: https://github.com/openai/glow#download-datasets However, these might require some manual processing to extract the images (as they are stored as TFRecords). You could probably use this as inspiration: https://github.com/NVlabs/NVAE#set-up-file-paths-and-data