Ahmadzei's picture
added 3 more tables for large emb model
5fa1a76
[TFEncoderDecoderModel.from_pretrained] currently doesn't support initializing the model from a
pytorch checkpoint.