Unable to load weights from pytorch checkpoint file

[En-J-A]

When I run a code to train a model in google colab, I got this kind of error

File "/usr/local/lib/python3.7/dist-packages/transformers/modeling_utils.py", line 1286, in from_pretrained
    f"Unable to load weights from pytorch checkpoint file for '{pretrained_model_name_or_path}' "
OSError: Unable to load weights from pytorch checkpoint file for 'model_name' at '/root/.cache/huggingface/transformers/8e7b2a80ddcb5611b27d8c89e1e8e33a947e105415051402a22b9c8d7d1caeb0.e22331f3a065b885b30ae3dd1ff11ccaf7fbc444485f6eb07ef5e0138bca8b70'If you tried to load a PyTorch model from a TF 2.0 checkpoint, please set from_tf=True. 
[ ]

where 'model_name' is a model exist in huggingface

But , when I run the same code on my local machine this error doesn't occur and the weights have loaded.

If you know the reason or how can I solve this issue, please guide me?

[WissamAntoun]

model_name should be replace by the model path in the hugging face library for example model_name='aubmindlab/bert-base-arabertv02'

https://huggingface.co/models?filter=ar

[WissamAntoun]

can you copy paste the related code here? also which version of transformers you are using

[En-J-A]

transformers==2.8.0
tokenizers==0.5.2
torch==1.4.0

[En-J-A]

I am not using arabert , I used both Bart and t5 model , Does this error relate to the model ?

tokenizer = MBart50TokenizerFast.from_pretrained(args.model_path if test else 'facebook/mbart-large-50',force_download=True)
model =  MBartForConditionalGeneration.from_pretrained(args.model_path if test else 'facebook/mbart-large-50')

[WissamAntoun]

try using the latest version of transformers

[En-J-A]

try using the latest version of transformers

Ok. I updated all the libraries to the last versions and it works.
thank you very much