selective_distillation
selective_distillation copied to clipboard
LeslieOverfitting
Results
2
selective_distillation issues
Sort by
recently updated
recently updated
newest added
You guys mention use of shared vocab in the paper, while your code generates two different vocab in WMT'14 En-De task, can you clarify about that
@LeslieOverfitting , Would you please share which code files did you develop, I have found `selective_distillation/tree/main/fairseq/criterions/label_smoothed_cross_entropy.py` is the one that you develop the distillation strategy. What else? Thanks!
