Very inspiring work. Could you please make your accuracy evaluation code public? I hope to be able to reproduce your experiments to learn better. By the way, your part about parallel computing is very eye-opening to me, but I don't fully understand it. I wonder if you can point this out to me with the code. Thank you very much!
The evaluation code is already released. You can refer to here for training and evaluation.
For parallel computing, I guess you mean our mask merging strategy. We convert mask merging to token merging for efficient computing. You can refer to here for more details.
Very inspiring work. Could you please make your accuracy evaluation code public? I hope to be able to reproduce your experiments to learn better. By the way, your part about parallel computing is very eye-opening to me, but I don't fully understand it. I wonder if you can point this out to me with the code. Thank you very much!