ghost
commented
1 year ago to be honest, the code given by the author does not reproduce the results in the text at all, I don't know if this is intentional on the author's part or why?
Open raycaohmu opened 1 year ago
ghost
commented
1 year ago to be honest, the code given by the author does not reproduce the results in the text at all, I don't know if this is intentional on the author's part or why?
I cloned the repository and ran 1_train_prior_Transformer.py. Then I used the trained ckpt to calculate success rate just like Table.1 in your article. But I got 9%, which is far less than the value 25.4% mentioned in the original paper. When I used the ckpt file you provided in the repository, I indeed got success rate of about 25%. I just wondered why I got a value of 9% when I trained from scratch using 1_train_prior_Transformer.py. Is there any detailed settings I need to care about? Please help! Thank you so much!