Current training pipeline only supports instruction-tuning based on the alpaca template/dataset
The baize's pipeline supports both training to follow instructions and to behave like a chatbot.
It might be interesting to incorporate this logic/dataset into our training pipeline
Current training pipeline only supports instruction-tuning based on the alpaca template/dataset The baize's pipeline supports both training to follow instructions and to behave like a chatbot. It might be interesting to incorporate this logic/dataset into our training pipeline