Question about reproducing results with README settings #15
Description
Hello, thank you for sharing this great work! I tried running the experiments by following the settings described in the README (grad_checkpointing = False), but I noticed that the results are not as good as the reported ones.
For example:
• My trained model on stage2 achieves 0.63
• While the results mentioned by the authors are around 0.73
May I ask if there are additional configurations, parameter tuning, or environment requirements that I should be aware of? I tried to strictly follow the steps in the README, but I might have missed some important details.
Thank you very much for your time and help!