"ValueError: Please make sure to properly initialize your"... when setting LOGLEVEL=DEBUG

System Info

- `Accelerate` version: 0.14.0
- Platform: Linux-5.4.0-124-generic-x86_64-with-debian-buster-sid
- Python version: 3.7.13
- Numpy version: 1.21.5
- PyTorch version (GPU?): 1.13.0+cu117 (True)
- `Accelerate` default config:
	Not found

Traceback (most recent call last):
  File "/workspace/diffusers/examples/dreambooth/train_dreambooth.py", line 689, in <module>
    main(args)
  File "/workspace/diffusers/examples/dreambooth/train_dreambooth.py", line 346, in main
    accelerator = Accelerator(
  File "/opt/conda/envs/dreambooth/lib/python3.10/site-packages/accelerate/accelerator.py", line 203, in __init__
    trackers = filter_trackers(log_with, self.logging_dir)
  File "/opt/conda/envs/dreambooth/lib/python3.10/site-packages/accelerate/tracking.py", line 580, in filter_trackers
    logger.debug(f"{log_with}")
  File "/opt/conda/envs/dreambooth/lib/python3.10/logging/__init__.py", line 1835, in debug
    self.log(DEBUG, msg, *args, **kwargs)
  File "/opt/conda/envs/dreambooth/lib/python3.10/site-packages/accelerate/logging.py", line 47, in log
    if self.isEnabledFor(level) and self._should_log(main_process_only):
  File "/opt/conda/envs/dreambooth/lib/python3.10/site-packages/accelerate/logging.py", line 32, in _should_log
    state = AcceleratorState()
  File "/opt/conda/envs/dreambooth/lib/python3.10/site-packages/accelerate/state.py", line 78, in __init__
    raise ValueError(
ValueError: Please make sure to properly initialize your accelerator via `accelerator = Accelerator()` before using any functionality from the `accelerate` library.

Information

• The official example scripts
• My own modified scripts

Tasks

• One of the scripts in the examples/ folder of Accelerate or an officially supported no_trainer script in the examples folder of the transformers repo (such as run_no_trainer_glue.py)
• My own task or dataset (give details below)

Reproduction

In the train_dreambooth.py example from the diffusers repo:

• https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth.py#L26

Just before the logger = get_logger(__name__) line.

Patch the following code into the file + set the LOGLEVEL env var to DEBUG:

import logging
LOGLEVEL = os.environ.get('LOGLEVEL', 'WARNING').upper()
logging.basicConfig(level=LOGLEVEL)

_Originally posted by @0xdevalias in https://github.com/huggingface/accelerate/issues/834#issuecomment-1309708720_

Then run with something like:

LOGLEVEL=DEBUG conda run -n dreambooth --live-stream \
  accelerate launch train_dreambooth.py \
    --pretrained_model_name_or_path="$MODEL_NAME_OR_PATH"  \
    --instance_data_dir="$INSTANCE_DIR" \
    --class_data_dir="$CLASS_REGULARISATION_IMAGE_DIR" \
    --output_dir="$OUTPUT_DIR" \
    --with_prior_preservation --prior_loss_weight=1.0 \
    --instance_prompt="$INSTANCE_PROMPT" \
    --class_prompt="$CLASS_REGULARISATION_PROMPT" \
    --resolution=512 \
    --train_batch_size=1 \
    --gradient_accumulation_steps=2 --gradient_checkpointing \
    --use_8bit_adam \
    --learning_rate=5e-6 \
    --lr_scheduler="constant" \
    --lr_warmup_steps=0 \
    --num_class_images="$NUM_CLASS_REGULARISATION_IMAGES_TO_USE" \
    --max_train_steps="$MAX_TRAIN_STEPS" \
    --seed=$SEED

The error is being raised by some code inside accelerate, when we’re calling the Accelerator initialiser:

• https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth.py#L332-L337

The code that seems to be triggering the error is the following logger.debug:

• https://github.com/huggingface/accelerate/blob/main/src/accelerate/tracking.py#L580

Expected behavior

The accelerate library wouldn’t crash, and would correctly output DEBUG level logs.