microsoft / Oscar

Oscar and VinVL
MIT License
1.04k stars 252 forks source link

Fine tune on the dataset with larger word dictionary? #162

Open CCYChongyanChen opened 3 years ago

CCYChongyanChen commented 3 years ago

Hi there, Could I know how to fine-tune a new dataset with a larger word dictionary?

The current model has a 3129 size word/answer dictionary while ours expended the 3129 to 50085 (our dictionary's word index from 0 to 3129 are the same as the original dictionary)

I currently have this error: size mismatch for classifier.weight: copying a param with shape torch.Size([3129, 1024]) from checkpoint, the shape in current model is torch.Size([50085, 1024]). size mismatch for classifier.bias: copying a param with shape torch.Size([3129]) from checkpoint, the shape in current model is torch.Size([50085]).

Thanks in advance!