Overcoming Catastrophic Forgetting in Zero-Shot Cross-Lingual Generation
Paper • 2205.12647 • Published
How to use DKYoon/mt5-small-lm-adapt with Transformers:
# Load model directly
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
tokenizer = AutoTokenizer.from_pretrained("DKYoon/mt5-small-lm-adapt")
model = AutoModelForSeq2SeqLM.from_pretrained("DKYoon/mt5-small-lm-adapt")🤗 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
Authors: Tu Vu, Aditya Barua, Brian Lester, Daniel Cer, Mohit Iyyer, Noah Constant
PyTorch port of the original Flax checkpoint at Google/T5X repository.