ERROR when torch.load("pytoch_model.bin") downloaded from huggingface about bluebert HOT 4 CLOSED

I didn't find the error. I use transformers==3.4.0 and torch==1.7.0.

>>> tokenizer = AutoTokenizer.from_pretrained("bionlp/bluebert_pubmed_mimic_uncased_L-12_H-768_A-12")
Downloading: 100%|██████████████████████████████████████████████████████████████████████| 313/313 [00:00<00:00, 395kB/s]
Downloading: 100%|███████████████████████████████████████████████████████████████████| 232k/232k [00:00<00:00, 2.91MB/s]
>>> model = AutoModel.from_pretrained("bionlp/bluebert_pubmed_mimic_uncased_L-12_H-768_A-12")
Downloading: 100%|███████████████████████████████████████████████████████████████████| 441M/441M [01:11<00:00, 6.17MB/s]
>>>

from bluebert.

I think it is a bug in the transformers package: huggingface/transformers#2420

from bluebert.

huggingface/transformers#2420

I think it is a bug in the transformers package: huggingface/transformers#2420

I don't think it's a problem with transformers package. I tried to use "torch.load" directly as follows but also got the same ERROR.

import torch
data = torch.load("pytoch_model.bin")

from bluebert.

I didn't find the error. I use transformers==3.4.0 and torch==1.7.0.

>>> tokenizer = AutoTokenizer.from_pretrained("bionlp/bluebert_pubmed_mimic_uncased_L-12_H-768_A-12")
Downloading: 100%|██████████████████████████████████████████████████████████████████████| 313/313 [00:00<00:00, 395kB/s]
Downloading: 100%|███████████████████████████████████████████████████████████████████| 232k/232k [00:00<00:00, 2.91MB/s]
>>> model = AutoModel.from_pretrained("bionlp/bluebert_pubmed_mimic_uncased_L-12_H-768_A-12")
Downloading: 100%|███████████████████████████████████████████████████████████████████| 441M/441M [01:11<00:00, 6.17MB/s]
>>>

GOT IT. It is the problem of the pytorch version
I could load the file successfully using torch=1.7.0
Thank you soooo much!

from bluebert.