Open glample opened 8 years ago
A good test case for this is to try running a vanilla RNN with -seq_length 1. When I try those settings on Enwik8, the difference in loss is quite noticeable (i.e. higher than 4 without the fix, below 2 with the fix).
I think a clone_list is missing. It gives different results without it (even if it is probably insignificant).