The steps for changing the language model are as follows:
- Find the huggingface name of the embeddings you want to use: https://huggingface.co/models. If you want to use a local model, you can just use the path to the embeddings as name
- Change the transformer model name (
transformer_model
) inconfigs/params.json
correspondingly (or make your own hyperparameters file. - That's it!, now train your model as normal