Hello, I really thank you for sharing your work.
I have some questions about your code.
I found that there are differences between the rendered image before and after saving the checkpoint.
Before saving checkpoint means that when I trained with opt.py, before saving checkpoint, I rendered the output image and calculated the metric.
After saving checkpoint means that during training I saved model checkpoint, and after training, I loaded checkpoint and rendered the output image, and calculate metrics.
When loading ckpt, rendered images became different (e.g., brighter locally, somewhat noisy).
Have you ever found this problem beforehand? If so, how did you solve it?
Hello, I really thank you for sharing your work. I have some questions about your code.
I found that there are differences between the rendered image before and after saving the checkpoint. Before saving checkpoint means that when I trained with
opt.py
, before saving checkpoint, I rendered the output image and calculated the metric.After saving checkpoint means that during training I saved model checkpoint, and after training, I loaded checkpoint and rendered the output image, and calculate metrics.
When loading ckpt, rendered images became different (e.g., brighter locally, somewhat noisy).
Have you ever found this problem beforehand? If so, how did you solve it?