Thanks for your implementation! I found something weird when using this training script, I hope u can help me to fix this problem.
I added test scripts during training to better visualize the training process. However, I found that the model could totally fail for some certain test cases shown as follows:
while perform normally for others:
And for the failed cases, the results are not getting better for further training. Is it normal? Have u ever met such similar circumstances?
Thanks for your implementation! I found something weird when using this training script, I hope u can help me to fix this problem. I added test scripts during training to better visualize the training process. However, I found that the model could totally fail for some certain test cases shown as follows: while perform normally for others:
And for the failed cases, the results are not getting better for further training. Is it normal? Have u ever met such similar circumstances?