Open LT1st opened 2 months ago
Change --eval_epoch_freq to any larger number of epoch (for e.g 100, it will run and throw error at 100th epoch). Its not a solution of the given problem, but a work around. You can train your model to as many number of epochs and then evaluate using test.py.
How to handle this problem?