Hyperparameters

[JiaqiLi404]

Hi

Thanks for your efforts! May I ask the eventual hyperparameters of the model and for training? Since I ran on the default seeting, the performance is less exciting. And can I know when the pretrained weights could be released?

Thanks!

[lsy0882]

We are currently re-running our research using architectural insights gained from the Mamba-v2 network and RDFA-S4, so the pretrained weights file has been temporarily removed.

Until the results of this new research are released, it might be beneficial to apply the recurrent operations method described in the paper to the Mamba-base network for the TAL task.