[Bug]: module 'torch' has no attribute 'float8_e4m3fn'

[sz1kormar]

Checklist

• The issue exists after disabling all extensions
• The issue exists on a clean installation of webui
• The issue is caused by an extension, but I believe it is caused by a bug in the webui
• The issue exists in the current version of the webui
• The issue has not been reported before recently
• The issue has been reported before but has not been fixed yet

What happened?

I was installing v1.9.3-amd as it was the best version for my RX 580 to run image generation with the fastest speed when using directml as newer versions slowing it down significantly.

It worked fine for a clean install for a long time until a strange bug happened

Steps to reproduce the problem

1. start webui-user.bat
2. wait for the installation to finish
3. the launch process begins
4. torch complains about fp8 even though I don't use flux just SD 1.5 and SDXL with fp16

What should have happened?

At runtime it should run without any issues at the start

What browsers do you use to access the UI ?

Other

Sysinfo

{
"Platform": "Windows-10-10.0.19045-SP0",
"Python": "3.10.6",
"Version": "v1.9.3-amd",
"Commit": "6130ef9e7a194bb0a212d12a697347153c4f31e2",
"Script path": "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu",
"Data path": "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu",
"Extensions dir": "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\extensions",
"Checksum": "88a9666aa54568c4b1b246753dcb06557a0ce1747983d5bb9b6355a6715e2ff4",
"Commandline": [
"launch.py",
"--use-directml",
"--skip-ort",
"--opt-sdp-attention",
"--medvram",
"--medvram-sdxl",
"--do-not-download-clip",
"--no-download-sd-model",
"--dump-sysinfo"
],
"Torch env info": {
"torch_version": "2.0.0+cpu",
"is_debug_build": "False",
"cuda_compiled_version": null,
"gcc_version": null,
"clang_version": null,
"cmake_version": null,
"os": "Microsoft Windows 10 Pro",
"libc_version": "N/A",
"python_version": "3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)] (64-bit runtime)",
"python_platform": "Windows-10-10.0.19045-SP0",
"is_cuda_available": "False",
"cuda_runtime_version": null,
"cuda_module_loading": "N/A",
"nvidia_driver_version": null,
"nvidia_gpu_models": null,
"cudnn_version": null,
"pip_version": "pip3",
"pip_packages": [
"numpy==1.26.2",
"open-clip-torch==2.20.0",
"pytorch-lightning==1.9.4",
"torch==2.0.0",
"torch-directml==0.2.0.dev230426",
"torchdiffeq==0.2.3",
"torchmetrics==1.6.1",
"torchsde==0.2.6",
"torchvision==0.15.1"
],
"conda_packages": null,
"hip_compiled_version": "N/A",
"hip_runtime_version": "N/A",
"miopen_runtime_version": "N/A",
"caching_allocator_config": "",
"is_xnnpack_available": "True",
"cpu_info": [
"Architecture=9",
"CurrentClockSpeed=3600",
"DeviceID=CPU0",
"Family=107",
"L2CacheSize=3072",
"L2CacheSpeed=",
"Manufacturer=AuthenticAMD",
"MaxClockSpeed=3600",
"Name=AMD Ryzen 5 3600 6-Core Processor ",
"ProcessorType=3",
"Revision=28928"
]
},
"Exceptions": [],
"CPU": {
"model": "AMD64 Family 23 Model 113 Stepping 0, AuthenticAMD",
"count logical": 12,
"count physical": 6
},
"RAM": {
"total": "16GB",
"used": "5GB",
"free": "11GB"
},
"GPU": "DirectML is not initialized",
"Extensions": [],
"Inactive extensions": [],
"Environment": {
"COMMANDLINE_ARGS": "--use-directml --skip-ort --opt-sdp-attention --medvram --medvram-sdxl --do-not-download-clip --no-download-sd-model --dump-sysinfo",
"GRADIO_ANALYTICS_ENABLED": "False"
},
"Config": "'NoneType' object has no attribute 'data'",
"Startup": null,
"Packages": [
"accelerate==0.21.0",
"aenum==3.1.15",
"aiofiles==23.2.1",
"aiohappyeyeballs==2.4.6",
"aiohttp==3.11.12",
"aiosignal==1.3.2",
"alembic==1.14.1",
"altair==5.5.0",
"antlr4-python3-runtime==4.9.3",
"anyio==3.7.1",
"async-timeout==5.0.1",
"attrs==25.1.0",
"blendmodes==2022",
"certifi==2025.1.31",
"charset-normalizer==3.4.1",
"clean-fid==0.1.35",
"click==8.1.8",
"clip==1.0",
"colorama==0.4.6",
"coloredlogs==15.0.1",
"colorlog==6.9.0",
"contourpy==1.3.1",
"cycler==0.12.1",
"deprecation==2.1.0",
"diffusers==0.32.2",
"diskcache==5.6.3",
"einops==0.4.1",
"exceptiongroup==1.2.2",
"facexlib==0.3.0",
"fastapi==0.94.0",
"ffmpy==0.5.0",
"filelock==3.17.0",
"filterpy==1.4.5",
"flatbuffers==25.2.10",
"fonttools==4.56.0",
"frozenlist==1.5.0",
"fsspec==2025.2.0",
"ftfy==6.3.1",
"gitdb==4.0.12",
"gitpython==3.1.32",
"gradio-client==0.5.0",
"gradio==3.41.2",
"greenlet==3.1.1",
"h11==0.12.0",
"httpcore==0.15.0",
"httpx==0.24.1",
"huggingface-hub==0.29.1",
"humanfriendly==10.0",
"idna==3.10",
"imageio==2.37.0",
"importlib-metadata==8.6.1",
"importlib-resources==6.5.2",
"inflection==0.5.1",
"jinja2==3.1.5",
"jsonmerge==1.8.0",
"jsonschema-specifications==2024.10.1",
"jsonschema==4.23.0",
"kiwisolver==1.4.8",
"kornia==0.6.7",
"lark==1.1.2",
"lazy-loader==0.4",
"lightning-utilities==0.12.0",
"llvmlite==0.44.0",
"mako==1.3.9",
"markupsafe==2.1.5",
"matplotlib==3.10.0",
"mpmath==1.3.0",
"multidict==6.1.0",
"narwhals==1.27.1",
"networkx==3.4.2",
"numba==0.61.0",
"numpy==1.26.2",
"olive-ai==0.7.1.1",
"omegaconf==2.2.3",
"onnx==1.17.0",
"onnxruntime==1.20.1",
"open-clip-torch==2.20.0",
"opencv-python==4.11.0.86",
"optimum==1.24.0",
"optuna==4.2.1",
"orjson==3.10.15",
"packaging==24.2",
"pandas==2.2.3",
"piexif==1.1.3",
"pillow-avif-plugin==1.4.3",
"pillow==9.5.0",
"pip==22.2.1",
"propcache==0.3.0",
"protobuf==3.20.3",
"psutil==5.9.5",
"pydantic==1.10.21",
"pydub==0.25.1",
"pyparsing==3.2.1",
"pyreadline3==3.5.4",
"python-dateutil==2.9.0.post0",
"python-multipart==0.0.20",
"pytorch-lightning==1.9.4",
"pytz==2025.1",
"pywavelets==1.8.0",
"pyyaml==6.0.2",
"referencing==0.36.2",
"regex==2024.11.6",
"requests==2.32.3",
"resize-right==0.0.2",
"rpds-py==0.23.1",
"safetensors==0.4.2",
"scikit-image==0.21.0",
"scipy==1.15.2",
"semantic-version==2.10.0",
"sentencepiece==0.2.0",
"setuptools==63.2.0",
"six==1.17.0",
"smmap==5.0.2",
"sniffio==1.3.1",
"spandrel==0.1.6",
"sqlalchemy==2.0.38",
"starlette==0.26.1",
"sympy==1.13.3",
"tifffile==2025.2.18",
"timm==1.0.14",
"tokenizers==0.13.3",
"tomesd==0.1.3",
"torch-directml==0.2.0.dev230426",
"torch==2.0.0",
"torchdiffeq==0.2.3",
"torchmetrics==1.6.1",
"torchsde==0.2.6",
"torchvision==0.15.1",
"tqdm==4.67.1",
"trampoline==0.1.2",
"transformers==4.30.2",
"typing-extensions==4.12.2",
"tzdata==2025.1",
"urllib3==2.3.0",
"uvicorn==0.34.0",
"wcwidth==0.2.13",
"websockets==11.0.3",
"yarl==1.18.3",
"zipp==3.21.0"
]
}

Console logs

venv "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug  1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.9.3-amd
Commit hash: 6130ef9e7a194bb0a212d12a697347153c4f31e2
Skipping onnxruntime installation.
C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\timm\models\layers\__init__.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
  warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers", FutureWarning)
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: `pytorch_lightning.utilities.distributed.rank_zero_only` has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from `pytorch_lightning.utilities` instead.
  rank_zero_deprecation(
Launching Web UI with arguments: --use-directml --skip-ort --opt-sdp-attention --medvram --medvram-sdxl --do-not-download-clip --no-download-sd-model
==============================================================================
You are running torch 2.0.0+cpu.
The program is tested to work with torch 2.1.2.
To reinstall the desired version, run with commandline flag --reinstall-torch.
Beware that this will cause a lot of large files to be downloaded, as well as
there are reports of issues with training tab on the latest version.

Use --skip-version-check commandline argument to disable this check.
==============================================================================
Calculating sha256 for C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\models\Stable-diffusion\illustriousSNAFUSuper_ilxlSNAFUV10.safetensors: Traceback (most recent call last):
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\utils\import_utils.py", line 920, in _get_module
    return importlib.import_module("." + module_name, self.__name__)
  File "C:\Users\Martin\AppData\Local\Programs\Python\Python310\lib\importlib\__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1050, in _gcd_import
  File "<frozen importlib._bootstrap>", line 1027, in _find_and_load
  File "<frozen importlib._bootstrap>", line 992, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
  File "<frozen importlib._bootstrap>", line 1050, in _gcd_import
  File "<frozen importlib._bootstrap>", line 1027, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1006, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 688, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 883, in exec_module
  File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\models\autoencoders\__init__.py", line 1, in <module>
    from .autoencoder_asym_kl import AsymmetricAutoencoderKL
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\models\autoencoders\autoencoder_asym_kl.py", line 22, in <module>
    from ..modeling_utils import ModelMixin
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\models\modeling_utils.py", line 35, in <module>
    from ..quantizers import DiffusersAutoQuantizer, DiffusersQuantizer
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\quantizers\__init__.py", line 15, in <module>
    from .auto import DiffusersAutoQuantizer
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\quantizers\auto.py", line 31, in <module>
    from .torchao import TorchAoHfQuantizer
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\quantizers\torchao\__init__.py", line 15, in <module>
    from .torchao_quantizer import TorchAoHfQuantizer
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\quantizers\torchao\torchao_quantizer.py", line 57, in <module>
    torch.float8_e4m3fn,
AttributeError: module 'torch' has no attribute 'float8_e4m3fn'

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

Traceback (most recent call last):
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\utils\import_utils.py", line 920, in _get_module
    return importlib.import_module("." + module_name, self.__name__)
  File "C:\Users\Martin\AppData\Local\Programs\Python\Python310\lib\importlib\__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1050, in _gcd_import
  File "<frozen importlib._bootstrap>", line 1027, in _find_and_load
  File "<frozen importlib._bootstrap>", line 1006, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 688, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 883, in exec_module
  File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\pipelines\stable_diffusion\pipeline_onnx_stable_diffusion.py", line 26, in <module>
    from ..pipeline_utils import DiffusionPipeline
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\pipelines\pipeline_utils.py", line 45, in <module>
    from ..models import AutoencoderKL
  File "<frozen importlib._bootstrap>", line 1075, in _handle_fromlist
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\utils\import_utils.py", line 910, in __getattr__
    module = self._get_module(self._class_to_module[name])
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\utils\import_utils.py", line 922, in _get_module
    raise RuntimeError(
RuntimeError: Failed to import diffusers.models.autoencoders.autoencoder_kl because of the following error (look up to see its traceback):
module 'torch' has no attribute 'float8_e4m3fn'

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

Traceback (most recent call last):
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\launch.py", line 48, in <module>
    main()
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\launch.py", line 44, in main
    start()
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\modules\launch_utils.py", line 701, in start
    webui.webui()
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\webui.py", line 64, in webui
    shared.demo = ui.create_ui()
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\modules\ui.py", line 1681, in create_ui
    settings.create_ui(loadsave, dummy_component)
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\modules\ui_settings.py", line 161, in create_ui
    component = create_setting_component(k)
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\modules\ui_settings.py", line 29, in create_setting_component
    args = info.component_args() if callable(info.component_args) else info.component_args
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\modules\shared_options.py", line 216, in <lambda>
    "diffusers_pipeline": OptionInfo('ONNX Stable Diffusion', 'Diffusers pipeline', gr.Dropdown, lambda: {"choices": list(shared_items.get_pipelines())}),
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\modules\shared_items.py", line 97, in get_pipelines
    'ONNX Stable Diffusion': getattr(diffusers, 'OnnxStableDiffusionPipeline', None),
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\utils\import_utils.py", line 911, in __getattr__
    value = getattr(module, name)
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\utils\import_utils.py", line 911, in __getattr__
    value = getattr(module, name)
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\utils\import_utils.py", line 910, in __getattr__
    module = self._get_module(self._class_to_module[name])
  File "C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\diffusers\utils\import_utils.py", line 922, in _get_module
    raise RuntimeError(
RuntimeError: Failed to import diffusers.pipelines.stable_diffusion.pipeline_onnx_stable_diffusion because of the following error (look up to see its traceback):
Failed to import diffusers.models.autoencoders.autoencoder_kl because of the following error (look up to see its traceback):
module 'torch' has no attribute 'float8_e4m3fn'
91fea33f81d1e7a25c63bd587def3c92d2c7a912dc3bf722705c5e5ce364cd21
Loading weights [91fea33f81] from C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\models\Stable-diffusion\illustriousSNAFUSuper_ilxlSNAFUV10.safetensors
Creating model from config: C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\repositories\generative-models\configs\inference\sd_xl_base.yaml
C:\Users\Martin\Downloads\stable-diffusion-webui-amdgpu\venv\lib\site-packages\huggingface_hub\file_download.py:797: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`.
  warnings.warn(
Applying attention optimization: sdp... done.
Model loaded in 30.9s (calculate hash: 20.1s, create model: 0.5s, apply weights to model: 8.5s, apply half(): 0.2s, hijack: 0.2s, calculate empty prompt: 1.2s).
Press any key to continue . . .

Additional information

• AMD driver WHQL 24.9.1
• Windows 10 19045 build
• Opera GX 116.0.5366.148

[sz1kormar]

After some research i managed to find a version that can handle the error and getting the generation around the same time or lower

original requirements (before bug / during bug):
torch==2.0.0
torchvision==0.15.1
torch-directml=0.2.0dev

fixed requirements
torch==2.2.1
torchvision==0.17.1
torch-directml==0.2.1.dev240521

Generation time for SDXL Illustrious:
~ 1 minute 20 seconds (Original requirements)
~ 1 minute 15 seconds (Fixed requirements)

Tests done with RX 580 8GB on v1.9.3-amd branch with these args:
--use-directml --skip-ort --opt-sdp-attention --medvram --medvram-sdxl --do-not-download-clip --no-download-sd-model