How to obtain frame_gt and frame_gt_mask in training .pt files?

[luyaohan666]

Hello,

I am trying to understand how the frame_gt and frame_gt_mask tensors are generated in the training .pt files. Could you clarify the process or point me to the relevant code section?

Thanks!

[luyaohan666]

@meneshail

[meneshail]

Hi,
Thank you for your interest!

I’d like to share more details on how the processing works. This part of the data processing flow is adapted from OpenFold, meaning that frame_gt and frame_gt_mask correspond to OpenFold’s rigidgroups_gt_frames and rigidgroups_gt_exists for the C, CA, and N frames. You should be able to implement this by referring to the OpenFold code.

I'm sorry that I don’t currently have the time to organize this portion of the code and documentation. However, I should be able to update this part in the next few weeks, and I'll be sure to let you know then.

[luyaohan666]

Hi,
Thank you so much!

I'll definitely check out the OpenFold code. I appreciate your help and look forward to the updated documentation.

[meneshail]

Hi @luyaohan666 ,

Just wanted to follow up on this issue.

I've now pushed the complete data preprocessing script to the repository. This new pipeline handles the full conversion from raw PDB files to the training features, including the generation of the frame_gt and frame_gt_mask tensors you asked about.

You can find instructions on how to run the script in the README.md under the Training > Data Preprocessing (Optional) section.

Thanks again for your interest and feel free to follow up for further questions.