How do I load the pretrained model? about spanbert HOT 4 CLOSED

Not quite sure what the setup is but torch.load is going to need a checkpoint. I think you'll have to untar the file and try with the checkpoint file.

from spanbert.

When I untar it, there are only two files, one is a config.json file, the other one is pytorch_model.bin. I used torch.load('pytorch_model.bin') but it gave me a collections.OrderedDict object, instead of a model object. Did I load it in the wrong way? Thanks

from spanbert.

That's expected. Perhaps, this might be useful:
https://github.com/facebookresearch/SpanBERT/blob/master/code/pytorch_pretrained_bert/modeling.py#L530

from spanbert.

Thank you. It gave me an empty BertPreTrainedModel() with this method

from spanbert.