Fine tune T5 for text generation in specific domain

Beginners
1

1 , I want to continue training a T5 model in huggingface on my own corpus ( about a specific domain)
2, Then I want to fine tune this model for text generation
I am worried that the model has a conflict between the 2 steps.
So, Is this possible to do?

2

I’m not sure what you mean. Training a model on a corpus is fine-turning a model. Make sure you choose a model well suited to the task you want your model to perform. For text generation you should start with a causal language model (choose text generation on the models page). Then I suggest making sure your corpus is well formatted so you don’t get anything you don’t want in your generations. The hardest part is managing the server resources for training.

3

Yes it is possible and is pretty easy now with the help of the HF team.

here are some resources and I guess do search for more on youtube or the HF blogs.

4