I'm trying to finetune an LLM with LoRA, using transformers' Trainer. In twenty minutes of training it didn't output any logs to screen or to disk, even though I set logging_steps to 1:
trainer = transformers.Trainer(
...
args = transformers.TrainingArguments(
...
logging_steps=1,
logging_dir="logs",
),
)
What do I need to do to see the training log? Set up any callbacks?