I have some questions in your paper and code
fsfwl11406 opened this issue · 1 comments
fsfwl11406 commented
Which code corresponds to Stage 4 in your paper ?Anneal_mode in your hyperparams is "cosine",but your code only use mode="hard",and where you give num_warmup_steps assignment.
congxin95 commented
About Stage 4, you can refer Step.3
in Usage
part in README.md
. Following the command, you can get the corresponding codes.
The anneal_mode
is set as "cosine" in hyperparams.py
. All of hyperparameters are set in hyperparams.py
actually. Please check it. Thank you.