I’m currently playing around with this model:
As you can see here, there’s a 2.5GB checkpoint file:
However, when I try to load the model, it doesn’t download the 2.5GB checkpoint and later complains that some of the weights were not used:
If I import the model a different way instead of using the pipeline factory method, I still have the same issue:
In both cases, it looks like the checkpoint file is being ignored and not being downloaded.
I guess it’s not an issue if the model is behaving as expected when I use it for paraphrasing text. However, in many cases, the quality of the paraphrased text that comes out is pretty poor (a lot of repeats, sometimes it simply outputs the exact same text that I give it as input, etc.)
So my question is:
- Do the warnings here mean anything or can I safely ignore them for my purposes?
- What’s the significance of that 2.5 GB checkpoint file? How do I utilize it?