Getting an error:
Error: You must call wandb.init() before wandb.log()
“Main NLP Tasks” in section “Fine tuning a Masked Language Model” at the following line
trainer.train()
The solution is to update the training arguments defined a few lines above.
training_args = TrainingArguments(
output_dir=f"{model_name}-finetuned-imdb",
overwrite_output_dir=True,
evaluation_strategy="epoch",
learning_rate=2e-5,
weight_decay=0.01,
per_device_train_batch_size=batch_size,
per_device_eval_batch_size=batch_size,
push_to_hub=True,
fp16=True,
logging_steps=logging_steps,
report_to="none" # ADD THIS LINE!
)
This is covered in another thread but an initial non-working answer is given, with another poster correcting None to “none” in a response. Also, other alternative solutions are offered that could be tried.