Missing the metadata_one.csv

[LeoHeTurnstone]

Hi Alexander, your work on FoldFlow is impressive. When I tried to run the training part using the command: python runner/train.py local=example, the program shown the error: [Errno 2] No such file or directory: './data/metadata_one.csv' And I can't find the file metadata_one.csv in ./data folder, could you kindly add the example file to the reporsitory? Thank you.

[ZichongWang]

+1, i also encounter this problem. Besides, the config file of wandb is also missing. It would be great if you can provide the missing file.

[stanislav-chekmenev]

Hi Alex! Thank you for the amazing work done. I'm also joining @LeoHeTurnstone and @ZichongWang on that issue. It would be great if you could provide the aforementioned file. Besides, are there any plans to provide the link to the full dataset that you used for training?

[ZichongWang]

We have successfully reproduced the results of FoldFlow, as a baseline of our method. We will release our code and all baseline method codes soon. In fact, the csv can be obtained via guidance of FrameDiff, but download ~80GB files is required. Alternatively, FrameFlow provides a processed data, and you can easily download from there. @stanislav-chekmenev

[stanislav-chekmenev]

@ZichongWang, thanks a lot! I really appreciate your help.

[OliviaViessmann]

I ran into the same issue. This is how I resolved it:

1. The code to pre-process a single or a directory of cifs/pdbs is in the FrameDiff repo here: https://github.com/jasonkyuyim/se3_diffusion/tree/master
2. I copied the script process_pdb_dataset.py into the FoldFlow repo I had cloned. This script includes a set of code in the FrameDiff data directory -- I copied over those as well. (alternatively you can of course clone the entire FrameFlow repo and run everything in there)
3. Run python process_pdb_dataset.py --mmcif_dir <path_to_my_cif_dir> . This created the pickle file of the pre-processed cif + the metadata.csv file that is required by the FrameFlow training scripts.