Logger is not defined

[teleprint-me]

Describe the bug

diffusers.quantizers.pipe_quant_config references an undefined logger instance.

Reproduction

Install torch, torchvision, and torchao nightly, then import PipelineQuantizationConfig, TorchAoConfig from diffusers.

Logs

Traceback (most recent call last):
  File ".venv/lib/python3.14/site-packages/diffusers/quantizers/torchao/torchao_quantizer.py", line 90, in _update_torch_safe_globals
    from torchao.dtypes.uintx.uint4_layout import UInt4Tensor
ModuleNotFoundError: No module named 'torchao.dtypes.uintx.uint4_layout'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File ".venv/lib/python3.14/site-packages/diffusers/utils/import_utils.py", line 1016, in _get_module
    return importlib.import_module("." + module_name, self.__name__)
           ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.14/importlib/__init__.py", line 88, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
           ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<frozen importlib._bootstrap>", line 1398, in _gcd_import
  File "<frozen importlib._bootstrap>", line 1371, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1314, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 491, in _call_with_frames_removed
  File "<frozen importlib._bootstrap>", line 1398, in _gcd_import
  File "<frozen importlib._bootstrap>", line 1371, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1342, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 938, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 759, in exec_module
  File "<frozen importlib._bootstrap>", line 491, in _call_with_frames_removed
  File ".venv/lib/python3.14/site-packages/diffusers/quantizers/__init__.py", line 16, in <module>
    from .auto import DiffusersAutoQuantizer
  File ".venv/lib/python3.14/site-packages/diffusers/quantizers/auto.py", line 35, in <module>
    from .torchao import TorchAoHfQuantizer
  File ".venv/lib/python3.14/site-packages/diffusers/quantizers/torchao/__init__.py", line 15, in <module>
    from .torchao_quantizer import TorchAoHfQuantizer
  File ".venv/lib/python3.14/site-packages/diffusers/quantizers/torchao/torchao_quantizer.py", line 111, in <module>
    _update_torch_safe_globals()
    ~~~~~~~~~~~~~~~~~~~~~~~~~~^^
  File ".venv/lib/python3.14/site-packages/diffusers/quantizers/torchao/torchao_quantizer.py", line 96, in _update_torch_safe_globals
    logger.warning(
    ^^^^^^
NameError: name 'logger' is not defined

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "txt2img.py", line 14, in <module>
    from diffusers import PipelineQuantizationConfig, TorchAoConfig, ZImagePipeline
  File "<frozen importlib._bootstrap>", line 1423, in _handle_fromlist
  File ".venv/lib/python3.14/site-packages/diffusers/utils/import_utils.py", line 1006, in __getattr__
    module = self._get_module(self._class_to_module[name])
  File ".venv/lib/python3.14/site-packages/diffusers/utils/import_utils.py", line 1018, in _get_module
    raise RuntimeError(
    ...<2 lines>...
    ) from e
RuntimeError: Failed to import diffusers.quantizers.pipe_quant_config because of the following error (look up to see its traceback):
name 'logger' is not defined

System Info

$ diffusers-cli env

Copy-and-paste the text below in your GitHub issue and FILL OUT the two last points.

- 🤗 Diffusers version: 0.36.0
- Platform: Linux-6.12.69-1-lts-x86_64-with-glibc2.42
- Running on Google Colab?: No
- Python version: 3.14.2
- PyTorch version (GPU?): 2.11.0.dev20260207+rocm7.1 (True)
- Flax version (CPU?/GPU?/TPU?): not installed (NA)
- Jax version: not installed
- JaxLib version: not installed
- Huggingface_hub version: 1.4.1
- Transformers version: 5.1.0
- Accelerate version: 1.12.0
- PEFT version: 0.18.1
- Bitsandbytes version: not installed
- Safetensors version: 0.7.0
- xFormers version: not installed
- Accelerator: NA
- Using GPU in script?: <fill in>
- Using distributed or parallel set-up in script?: <fill in>

Who can help?

I was able to resolve the issue by simply referencing the default logger.

from ...utils.logging import _get_library_root_logger

Then adding the following line after checking for torch and torchao.

logger = _get_library_root_logger()

I'm not sure if this is the right way to fix it, but it the issue is resolved after adding these 2 lines to diffusers/quantizers/torchao/torchao_quantizer.py.

This is a nightly release, so I expect issues to appear from time to time. This seems like a simple oversight and is easy to fix.

[teleprint-me]

I accidentally closed this issue.

I just wanted to add that torchao does not have the UInt4Tensor class definition nor the uint4_layout module.

This blocks quantization execution due to lazy loading.

        from torchao.dtypes import NF4Tensor
        from torchao.dtypes.floatx.float8_layout import Float8AQTTensorImpl
        # There is no uint4_layout module and there is no UInt4Tensor class
        # for torchao >= 0.15.0
        # from torchao.dtypes.uintx.uint4_layout import UInt4Tensor
        from torchao.dtypes.uintx.uintx_layout import UintxAQTTensorImpl, UintxTensor

        # Remove the UInt4Tensor reference so that the code can be executed
        # safe_globals.extend([UintxTensor, UInt4Tensor, UintxAQTTensorImpl, Float8AQTTensorImpl, NF4Tensor])
        safe_globals.extend([UintxTensor, UintxAQTTensorImpl, Float8AQTTensorImpl, NF4Tensor])

The exception ends up blocking what would otherwise execute just fine.

Batching the imports is messy. Each import can be handled individually, but a better path might be to handle them conditionally.

[cardamon]

It appears this is already been fixed on the main branch. A quick workaround is to simply edit your local copy of site-packages/diffusers/quantizers/torchao/torchao_quantizer.py and move the line at L126

logger = logging.get_logger(__name__)

up to around L74 (or higher)

[teleprint-me]

Yes, I see the logger has been updated, but the import for uint4 is still invalid.