Overfitting when training with 2 GPUs about eda HOT 2 CLOSED

Thank you for your interest in our work and your feedback.

• It seems that you only trained for 51 epochs. We recommend that you consider increasing the training time.
  For your reference, we have included our last two training records below.

  # 1 [4 * 3090 GPUs]
  epoch 51: 51.87
  epoch 57: 54.05
  epoch 75: 54.22
  # 2 [4 * 3090 GPUs]
  epoch 51: 51.35
  epoch 57: 53.99
  epoch 66: 54.44

  Please note that slight performance variances of about 1% may occur between our repeated attempts, particularly when using different environments and devices.
• The trend of the verification loss appears to be normal, which is consistent with our experiments.

from eda.

Close it since no response. Feel free to re-open it if problems still exist.

from eda.