Open ljjcoder opened 2 years ago
Before releasing the code, we refactored the code on a large scale, which brings mixed precision training and more concise logic. Therefore, it may be helpful to fine tune the hyperparameters and turn off the mixed precision. We may update these hyperparameter settings later to adapt to the latest code version.
Thanks for your excellent work! when I run the commands python train.py --config_file configs/mfa.yml -g 2. I only get 54.80 MIOU. Can you give me some advice to improve the performance to 55.7?