AttributeError: 'Trainer' object has no attribute 'is_world_master' while fine tuning BioBERT-PyTorch on my custom data set

[abhibisht89]

Environment info:
transformers version: 4.0.0
Platform: Linux-4.15.0-96-generic-x86_64-with-debian-buster-sid
Python version: 3.7.8
PyTorch version (GPU?): 1.7.0 (True)

while running run_ner.py , getting above error at below line:
if trainer.is_world_master():
tokenizer.save_pretrained(training_args.output_dir)

[ankan94]

Have you resolved this?
I'm getting the same error at this level. The is_world_master has probably been changed to is_world_process_zero. You can give it a try.

[abhibisht89]

@ankan94 yeah , I did the same thing replace "is_world_master" by "is_world_process_zero" in the code and it is working as expected for me

[abhibisht89]

found a work around as mention above

[wzkariampuzha]

Please fix this in the run_ner code @dmis-lab

[rkatriel]

This issue is still present in the current GitHub version. I followed the directions on the main web page (https://github.com/dmis-lab/biobert-pytorch) and got the same error:

Traceback (most recent call last):
File "/Users/ron/Documents/GitHub/biobert-pytorch-master/named-entity-recognition/run_ner.py", line 328, in
main()
File "/Users/ron/Documents/GitHub/biobert-pytorch-master/named-entity-recognition/run_ner.py", line 249, in main
if trainer.is_world_master():
^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'Trainer' object has no attribute 'is_world_master'

Replacing "is_world_master" by "is_world_process_zero", as suggested above, makes the error go away but it should be fixed at the source in Git.

In case it matters, this is on an MacBook Air (M2, 2022) with macOS 12.6 (Monterey).