huggingface / transformers

🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
https://huggingface.co/transformers
Apache License 2.0
133.31k stars 26.63k forks source link

UL2 Mixture-of-Denoiser loss #21208

Closed gaceladri closed 1 year ago

gaceladri commented 1 year ago

Feature request

The losses applied to the paper UL2: Unifying Language Learning Paradigms The Mixture-of-Denoisers losses are described in the UL2 paper, which can be found at the following link: https://arxiv.org/abs/2205.05131 The code is based on T5x (which is JAX/FLAX): https://github.com/google-research/t5x

Motivation

I am requesting the addition of new losses applied in the UL2 paper called Mixture-of-Denoisers. These new losses have been shown to improve the performance of unsupervised learning models and I believe they could benefit the HuggingFace community.

Your contribution

Opening the request

github-actions[bot] commented 1 year ago

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.