Open dypromise opened 1 year ago
Thank you for your great work!
I found that the default div_factor is 25 and final_div_factor is 1e4, and your code's --lr=2e-4, so the learning rate when finished training is lr / 25 / 1e4 = 8e-10? Is that true? I think it is too small
Best regards.
Thank you for your great work!
I found that the default div_factor is 25 and final_div_factor is 1e4, and your code's --lr=2e-4, so the learning rate when finished training is lr / 25 / 1e4 = 8e-10? Is that true? I think it is too small
Best regards.