GPT2Model model output inconsistency between different transformers versions

After digging into it a little deeper, I found that the model output inconsistency was introduced between transformers v4.41.0 and v4.42.0.

1 Like