skip to Main Content

I am trying to test this model called BETO (the model is an implementation of Bert in Spanish):

!pip install transformers
!wget https://users.dcc.uchile.cl/~jperez/beto/cased_2M/pytorch_weights.tar.gz 
!wget https://users.dcc.uchile.cl/~jperez/beto/cased_2M/vocab.txt 
!wget https://users.dcc.uchile.cl/~jperez/beto/cased_2M/config.json 
!tar -xzvf pytorch_weights.tar.gz
!mv config.json pytorch/.
!mv vocab.txt pytorch/.

import torch
from transformers import BertForMaskedLM, BertTokenizer
tokenizer = BertTokenizer.from_pretrained("pytorch/", do_lower_case=False)
model = BertForMaskedLM.from_pretrained("pytorch/")
model.eval()

the enviroment is as follows:

platform            debian 10
transformers            3.4.0
python                  3.7.3
torch                   1.7.0
tensorflow              2.3.1

but in the following line:

model = BertForMaskedLM.from_pretrained("pytorch/")

I get this error:

Exception has occurred: OSError
Unable to load weights from pytorch checkpoint file. If you tried to load a PyTorch model from a TF 2.0 checkpoint, please set from_tf=True.

Thanks in advance

2

Answers


  1. I’ve tried the exact same code in Ubuntu with no problems.

    OSError is a built-in exception in Python and serves as the error class for the os module, which is raised when an os specific system function returns a system-related error, including I/O failures such as “file not found” or “disk full”.

    Maybe you can dig deeper into your OSError?

    Login or Signup to reply.
  2. try with this:

    !pip install urllib3==1.25.10
    !git clone https://github.com/huggingface/transformers 
    && cd transformers 
    && git checkout a3085020ed0d81d4903c50967687192e3101e770
    
    !pip install transformers
    !wget https://users.dcc.uchile.cl/~jperez/beto/cased_2M/pytorch_weights.tar.gz 
    !wget https://users.dcc.uchile.cl/~jperez/beto/cased_2M/vocab.txt 
    !wget https://users.dcc.uchile.cl/~jperez/beto/cased_2M/config.json 
    !tar -xzvf pytorch_weights.tar.gz
    !mv config.json pytorch/.
    !mv vocab.txt pytorch/.
    
    tokenizer = BertTokenizer.from_pretrained("pytorch/", do_lower_case=False)
    
    Login or Signup to reply.
Please signup or login to give your own answer.
Back To Top
Search