Pytorch/Xformers incompatibilities prevent Invoke from starting

[FoxTrotte]

Hey !

I just tried installing Invoke through Pinokio on my Windows 11 machine equipped with an RTX 5090.

Unfortunately, after the install is done, starting Invoke gives me this error:

C:\pinokio\api\invoke.git\app>conda_hook && conda deactivate && conda deactivate && conda deactivate && conda activate base && C:\pinokio\api\invoke.git\app\env\Scripts\activate C:\pinokio\api\invoke.git\app\env && invokeai-web
WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
    PyTorch 2.7.1+cu126 with CUDA 1208 (you have 2.7.1+cpu)
    Python  3.9.13 (you have 3.10.16)
  Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers)
  Memory-efficient attention, SwiGLU, sparse and more won't be available.
  Set XFORMERS_MORE_DETAILS=1 for more details
Traceback (most recent call last):
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\diffusers\utils\import_utils.py", line 820, in _get_module
    return importlib.import_module("." + module_name, self.__name__)
  File "C:\pinokio\bin\miniconda\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:\pinokio\api\invoke.git\app\env\lib\site-packages\diffusers\models\autoencoders\__init__.py", line 1, in <module>
    from .autoencoder_asym_kl import AsymmetricAutoencoderKL
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\diffusers\models\autoencoders\autoencoder_asym_kl.py", line 23, in <module>
    from .vae import DecoderOutput, DiagonalGaussianDistribution, Encoder, MaskConditionDecoder
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\diffusers\models\autoencoders\vae.py", line 24, in <module>
    from ..attention_processor import SpatialNorm
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\diffusers\models\attention_processor.py", line 35, in <module>
    import xformers.ops
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\xformers\ops\__init__.py", line 9, in <module>
    from .fmha import (
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\xformers\ops\fmha\__init__.py", line 10, in <module>
    from . import (
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\xformers\ops\fmha\flash3.py", line 109, in <module>
    from ...flash_attn_3 import _C  # type: ignore[attr-defined]  # noqa: F401
ImportError: DLL load failed while importing _C: Le module spécifié est introuvable.

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

Traceback (most recent call last):
  File "C:\pinokio\bin\miniconda\lib\runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "C:\pinokio\bin\miniconda\lib\runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "C:\pinokio\api\invoke.git\app\env\Scripts\invokeai-web.exe\__main__.py", line 7, in <module>
    sys.exit(run_app())
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\invokeai\app\run_app.py", line 35, in run_app
    from invokeai.app.invocations.baseinvocation import InvocationRegistry
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\invokeai\app\invocations\baseinvocation.py", line 36, in <module>
    from invokeai.app.invocations.fields import (
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\invokeai\app\invocations\fields.py", line 10, in <module>
    from invokeai.backend.model_manager.taxonomy import (
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\invokeai\backend\model_manager\taxonomy.py", line 7, in <module>
    from diffusers.pipelines.pipeline_utils import DiffusionPipeline
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\diffusers\pipelines\pipeline_utils.py", line 47, in <module>
    from ..models import AutoencoderKL
  File "<frozen importlib._bootstrap>", line 1075, in _handle_fromlist
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\diffusers\utils\import_utils.py", line 810, in __getattr__
    module = self._get_module(self._class_to_module[name])
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\diffusers\utils\import_utils.py", line 822, 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):
DLL load failed while importing _C: Le module spécifié est introuvable.

Now I tried installing Pytorch 2.7.1+cu128 (cu126 is unsupported on RTX5000s apparently) but after doing so I'm getting the following error when launching Invoke:

Traceback (most recent call last):
  File "C:\pinokio\bin\miniconda\lib\runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "C:\pinokio\bin\miniconda\lib\runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "C:\pinokio\api\invoke.git\app\env\Scripts\invokeai-web.exe\__main__.py", line 7, in <module>
    sys.exit(run_app())
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\invokeai\app\run_app.py", line 35, in run_app
    from invokeai.app.invocations.baseinvocation import InvocationRegistry
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\invokeai\app\invocations\baseinvocation.py", line 44, in <module>
    from invokeai.app.services.shared.invocation_context import InvocationContext
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\invokeai\app\services\shared\invocation_context.py", line 18, in <module>
    from invokeai.app.services.model_records.model_records_base import UnknownModelException
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\invokeai\app\services\model_records\__init__.py", line 3, in <module>
    from .model_records_base import (  # noqa F401
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\invokeai\app\services\model_records\model_records_base.py", line 15, in <module>
    from invokeai.backend.model_manager.configs.controlnet import ControlAdapterDefaultSettings
  File "C:\pinokio\api\invoke.git\app\env\lib\site-packages\invokeai\backend\model_manager\configs\controlnet.py", line 1, in <module>
    from typing import (
ImportError: cannot import name 'Self' from 'typing' (C:\pinokio\bin\miniconda\lib\typing.py)

Maybe the issue also lies with the Python version used ?

I'm out of ideas here !

Thanks a lot!

[FoxTrotte]

Using this fork works wonders