add readme

README.md

@@ -1,3 +1,11 @@
 ---
 license: apache-2.0
 ---
+
+🤗 Language model initialized from mT5 and trained for an additional 100K steps on the Prefix LM objective using mC4 data.
+
+Paper: [Overcoming Catastrophic Forgetting in Zero-Shot Cross-Lingual Generation](https://arxiv.org/abs/2205.12647)
+
+Authors: Tu Vu, Aditya Barua, Brian Lester, Daniel Cer, Mohit Iyyer, Noah Constant
+
+PyTorch port of the original Flax checkpoint at [Google/T5X repository](https://github.com/google-research/t5x).