microsoft

/

harrier-oss-v1-0.6b

like 271

Follow

Microsoft 20.8k

Couldn't you have just trained this one on the same dataset from the 27b? If anything it made this model worse and cost more to train.