Dear authors,
Thank you very much for your excellent work on CoMA. I found your paper and project very inspiring and impressive.
I am currently studying compositional motion generation and would like to better understand and reproduce your training process. I noticed that the repository provides inference and demo scripts, but the training code does not seem to be included.
Would it be possible for you to share the training scripts (train_res_Timesformer.py, train_vq_general.py , train_vq_multi.py, train_t2m_Timesformer_recap.py), or provide any guidance on how to train the models from scratch?
I understand that preparing and maintaining training code can take time, so I completely understand if it's not available at the moment. Any help or pointers would be greatly appreciated.
Thank you again for your great contribution!
Best regards,
Jiawei
Dear authors,
Thank you very much for your excellent work on CoMA. I found your paper and project very inspiring and impressive.
I am currently studying compositional motion generation and would like to better understand and reproduce your training process. I noticed that the repository provides inference and demo scripts, but the training code does not seem to be included.
Would it be possible for you to share the training scripts (
train_res_Timesformer.py,train_vq_general.py,train_vq_multi.py,train_t2m_Timesformer_recap.py), or provide any guidance on how to train the models from scratch?I understand that preparing and maintaining training code can take time, so I completely understand if it's not available at the moment. Any help or pointers would be greatly appreciated.
Thank you again for your great contribution!
Best regards,
Jiawei