Request for Training Script to Reproduce Paper Results

[pandeymansi]

Hello,

First, I'd like to congratulate you on your excellent work on RobotDiffuse. The paper is very insightful, and the approach of using an encoder-only transformer for motion planning is fascinating.

I am currently working on reproducing the results presented in your paper and have clones the official GitHub repository. I've successfully located the model architecture, the dataset generation scripts, and other utilities, which are very helpful.

However, I seem to be unable to find the main training script that integrates the data loading, model instruction, loss calculation, and optimization loop.

Would it be possible for you to share or upload the training script you used for the paper? Having access to it would be immensely helpful for the research community to accurately reproduce your results and build upon your important work.

Thank you for your time and for your significant contribution to the field of robotic motion planning.

[ACRoboT-buaa]

Thank you for your interest in our work and for reaching out!

Unfortunately, as our paper was not accepted, we are currently in the process of refining the methodology and preparing a revised version for resubmission. For this reason, the training code is not yet publicly available.

We appreciate your understanding and will make the code open source if the updated work is accepted.