Huggingface Trainer throws an AttributeError:'Namespace' object has no attribute 'get_process_log_level

Question

I am trying to run Trainer from Hugging face(pytorch) with arguments parser. My code looks like

if __name__ == '__main__':
    parser = HfArgumentParser(TrainingArguments)
    parser.add_argument('--model_name_or_path', type=str, required=True)
    .
    .
    .   
    .


    training_args = parser.parse_args()

    print('args', training_args)

    os.makedirs(training_args.output_dir, exist_ok=True)

    random.seed(training_args.seed)
    
    set_seed(training_args.seed)

    dataset_train = ...

    .
    .
    .


    trainer = Trainer(
        model=model,
        args=training_args,
        train_dataset=train_dataloader,
        eval_dataset=val_dataloader)

    trainer.train()

I am getting the following error:

    Traceback (most recent call last):
    File "main.py", line 250, in <module>
    eval_dataset=val_dataloader)
    File "C:\User\transformer\lib\site- 
    packages\transformers\trainer.py", line 316, in __init__
    log_level = args.get_process_log_level()
    AttributeError: 'Namespace' object has no attribute 'get_process_log_level

Any idea about this error and how to solved it?

Answer 1

First check that this works for you:

from transformers import TrainingArguments

args = TrainingArguments(output_dir='./')
args.get_process_log_level()

[out]:

20

If it doesn't then most probably the version of transformers you have on C:\User\transformer\lib\site-packages\transformers doesn't match the Trainer script you have. Then try to upgrade your transformers version pip install -U transformers.

---

If you get the 20 output but when you run your script you're getting the error, then most probably the version of your transformers is from a previous version that doesn't have the get_process_log_level() in the TrainingArguments as a property.

Add this line at the top of your code and check that the

import sys; print(sys.executable)

you'll get something like

[out]:

C://something/somewhere/bin/python

and with that do this to upgrade the library to the right site-package location and python binary

C://something/somewhere/bin/python -m pip install -U transformers

After upgrading the script should run.