Can’t download the pre-trained pegasus-large model

[xiaohy9]

Environment info

• transformers version: 4.4.2
• Platform:
• Python version: 3.7.10
• PyTorch version (GPU?):
• Tensorflow version (GPU?): 2.4.1
• Using GPU in script?: yes
• Using distributed or parallel set-up in script?:

Who can help

• blenderbot, bart, marian, pegasus, encoderdecoder, t5: @patrickvonplaten, @patil-suraj

Information

It appears that the huggingface.co model url has some problem.

Code:

from transformers import AutoTokenizer, TFAutoModelForSeq2SeqLM
tokenizer2 = AutoTokenizer.from_pretrained("google/pegasus-large")  
model2 = TFAutoModelForSeq2SeqLM.from_pretrained("google/pegasus-large")
inputs1_2 = tokenizer2.encode("summarize: " + text1, return_tensors="tf", max_length=1024)
outputs1_2 = model2.generate(inputs1_2, max_length=150, min_length=40, length_penalty=2.0, num_beams=4, early_stopping=True)
outputs1_2, tokenizer2.decode(outputs1_2[0])

error message:

404 Client Error: Not Found for url: https://huggingface.co/google/pegasus-large/resolve/main/tf_model.h5

…

OSError: Can't load weights for 'google/pegasus-large'. Make sure that:

- 'google/pegasus-large' is a correct model identifier listed on 'https://huggingface.co/models'

- or 'google/pegasus-large' is the correct path to a directory containing a file named one of tf_model.h5, pytorch_model.bin.

[patrickvonplaten]

Hey @xiaohy9,

Thanks for the issue! It should be resolved now. See: https://huggingface.co/google/pegasus-large/commit/4510ba69cc183d23e892e7728a40fdcf42e83079 .

Could you try again?

[xiaohy9]

Yes, it works now. Thanks for the quick response!
However, I saw some similar issue using pegasus-large as using pegasus-xsum, with details mentioned here: #10837

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.