search

Shark-NLP / DiffuSeq

[ICLR'23] DiffuSeq: Sequence to Sequence Text Generation with Diffusion Models
MIT License
728 stars 88 forks source link

Sucessfully training log #58

Closed skepsun closed 1 year ago

skepsun commented 1 year ago

Thanks for your great work. I wonder how various loss curves (mse, decoder_nll, tT, nll) change in a sucessful training. Can you or anyone please share some wandb/tensorboard logs?

skepsun commented 1 year ago

Well I found one.