Hi,

May I know how did you approch the training process?

If you have two different domains, it is recommended to fine-tune using the base model

1) base_model -> fine-tune domain 1

2) base_model -> fine-tune domain 2

instead of

3) base_model -> domain1 -> domain2

Do let me know if this is not the case in your project as I am a bit confused on the issue that you faced.