Update checkpoint for transformers>=4.29
#5
by ArthurZ HF Staff - opened
ArthurZ changed pull request status to merged
Automatically merging the PR.
Following the merge of a PR in transformers it appeared that this model was not properly converted. This PR will fix the inference and was tested using the following script:
>>> from transformers import AutoTokenizer, MarianMTModel
>>> tokenizer = AutoTokenizer.from_pretrained('Helsinki-NLP/opus-tatoeba-en-ro')
>>> inputs = tokenizer("Hey! Let's learn together", return_tensors="pt", padding=True)
>>> model = MarianMTModel.from_pretrained('Helsinki-NLP/opus-tatoeba-en-ro')
>>> print(tokenizer.batch_decode(model.generate(**inputs)))
['<pad> Hai să învăţăm împreună.</s>']