HuBERT Large Reproduce

[jxiw]

We are trying to reproduce HuBERT large.

Could you please share some pretraining information, for example loss, perplexity, correct_m, correct_u? We are not sure where does our current model stand.

Here are my training curves. Are there any problems? Which loss do i need to reach?

Thanks.