POSTECH-CVLab / point-transformer

This is an unofficial implementation of the Point Transformer paper.
508 stars 99 forks source link

Is your Loss converged? #69

Open Avril-Dragon opened 1 year ago

Avril-Dragon commented 1 year ago

Thank you for your outstanding performance in this job. I am currently using a 3090 for training, and the batch size has been adjusted to 4. I have been training for 50 rounds, and the loss has been fluctuating without a downward trend. I would like to ask how your loss performance during the training process has been