#1747·StabilityMatrix

Issue: Wan2GP NVIDIA installer forces an obsolete Torch/CUDA stack and fails to install SageAttention under Python 3.11

Author: NerfatronCreated Sep 19, 2026Updated Sep 19, 2026

Labelsbug

Package

Wan2GP

When did the issue occur?

Installing the Package

What GPU / hardware type are you using?

Nvidia RTX 5060Ti 16GB

What happened?

Stability Matrix Wan2GP installer uses an obsolete NVIDIA dependency stack on RTX 50 / Python 3.11

Summary

Issue: Wan2GP NVIDIA installer forces an obsolete Torch/CUDA stack and fails to install SageAttention under Python 3.11

Environment

Stability Matrix
Wan2GP
Python 3.11.14
RTX 5060 Ti (Blackwell)

Problem

When creating a new Wan2GP package with Python 3.11.14, Stability Matrix initially resolves the current Wan2GP requirements successfully, including:

torch==2.14.0
torchvision==0.29.0
torchaudio==2.11.0

However, the Stability Matrix-specific NVIDIA installation step then explicitly replaces these with the much older:

torch==2.7.1+cu128
torchvision==0.22.1+cu128
torchaudio==2.7.1+cu128
xformers==0.0.30
triton-windows==3.3.1.post19

It subsequently attempts to install:

sageattention==2.2.0+cu128torch2.7.1

That installation fails because the available SageAttention wheel has no Python 3.11 (cp311) build:

Because only sageattention==2.2.0+cu128torch2.7.1 is available and
sageattention==2.2.0+cu128torch2.7.1 has no wheels with a matching
Python version tag (e.g., `cp311`), we can conclude that all versions
of sageattention cannot be used.

The traceback identifies the Stability Matrix code responsible as:

StabilityMatrix.Core.Models.Packages.Wan2GP.InstallNvidiaAsync

Why this matters

This appears to be more than a broken SageAttention dependency. The NVIDIA installer is hard-coded around an older CUDA 12.8 / PyTorch 2.7.1 configuration, rather than following Wan2GP's current NVIDIA/RTX 50 installation path.

For Blackwell/RTX 50 users, this has significant consequences. The forced legacy stack prevents the package from using the newer CUDA 13 / PyTorch 2.10-based ecosystem and consequently locks users out of newer optimisation paths such as:

modern SageAttention builds
NVFP4-related functionality
the newer RTX 50-specific Wan2GP optimisation stack
other attention/kernel packages that depend on the newer Torch/CUDA ABI

The irony is that the failure occurs after Stability Matrix has successfully created a Python 3.11 environment. The installer then selects an older SageAttention package that only has a CPython 3.10 wheel, making the installation inherently incompatible with the Python version it is using.

Reproduction

A fresh Wan2GP installation was created in Stability Matrix using Python 3.11.14. The same behaviour occurs regardless of selecting the Main or fixes Wan2GP branch.

The relevant installation sequence from the log is:

Python 3.11.14 environment
        ↓
Normal Wan2GP requirements resolve successfully
        ↓
Torch 2.14.0 / torchvision 0.29.0 / torchaudio 2.11.0
        ↓
InstallNvidiaAsync replaces them
        ↓
Torch 2.7.1+cu128 / torchvision 0.22.1+cu128 /
torchaudio 2.7.1+cu128 / xformers 0.0.30
        ↓
Triton 3.3.1.post19
        ↓
SageAttention 2.2.0+cu128torch2.7.1
        ↓
No CPython 3.11 wheel
        ↓
Installation fails

Requested fix

Please update InstallNvidiaAsync so that the NVIDIA installation path follows the current Wan2GP installation configuration rather than forcing the legacy Torch 2.7.1/cu128/SageAttention combination.

In particular, the installer should support the current RTX 50/Blackwell stack and its compatible Python 3.11, CUDA 13, Torch, Triton and SageAttention/NVFP4 components.

The existing working Wan2GP installation confirms that the problem is specifically with the Stability Matrix installation path; the issue is reproducible with a clean package and does not require modifying an existing installation.

Console output

Unpacking resources Cloning into 'C:\AI\StabilityMatrix\Data\Packages\Wan2GP - Python 3.11.14'... Download Complete Using Python 3.11.14 environment at: venv Resolved 3 packages in 138ms Prepared 2 packages in 33ms Installed 2 packages in 35ms

packaging==26.3
wheel==0.48.0 Using Python 3.11.14 environment at: venv Resolved 310 packages in 7.27s Downloading tokenizers (2.4MiB) Downloading babel (9.7MiB) Downloading scikit-learn (7.9MiB) Downloading symusic (2.1MiB) Downloading spacy-pkuseg (2.3MiB) Downloading torchcodec (2.1MiB) Downloading transformers (10.7MiB) Downloading virtualenv (6.8MiB) Downloading pywin32 (6.6MiB) Downloading numba (2.7MiB) Downloading fonttools (2.4MiB) Downloading cython (2.7MiB) Downloading sherpa-onnx-core (14.9MiB) Downloading polars-runtime-32 (48.9MiB) Downloading sherpa-onnx (2.1MiB) Downloading matplotlib (8.9MiB) Downloading timm (2.5MiB) Downloading pygame (10.1MiB) Downloading pandas (10.8MiB) Downloading sympy (6.0MiB) Downloading blis (5.9MiB) Downloading resampy (2.9MiB) Downloading cryptography (3.7MiB) Downloading numpy (12.3MiB) Downloading jax (3.1MiB) Downloading torch (118.3MiB) Downloading diffusers (4.4MiB) Downloading scipy (34.9MiB) Downloading misaki (3.5MiB) Downloading gradio (51.6MiB) Downloading samplerate (3.8MiB) Downloading taichi (79.4MiB) Downloading decord (23.6MiB) Downloading opencv-python (42.0MiB) Downloading onnx (7.5MiB) Downloading onnx-weekly (7.5MiB) Downloading jaxlib (62.8MiB) Downloading hf-xet (3.8MiB) Downloading opencv-python-headless (41.8MiB) Downloading pillow (6.7MiB) Downloading espeakng-loader (9.0MiB) Downloading spacy (15.6MiB) Downloading imageio-ffmpeg (29.8MiB) Downloading scikit-image (11.3MiB) Downloading ruff (10.1MiB) Downloading llvmlite (39.9MiB) Downloading onnxruntime-gpu (174.3MiB) Downloading sqlalchemy (2.1MiB) Downloading networkx (2.0MiB) Downloading dashscope (1.9MiB) Downloaded sherpa-onnx Downloaded symusic Downloading pydantic-core (1.9MiB) Downloading thinc (1.7MiB) Downloaded spacy-pkuseg Downloading wetext (1.7MiB) Downloaded torchcodec Downloading open-clip-torch (1.5MiB) Downloaded fonttools Downloaded tokenizers Downloading ultralytics (1.4MiB) Downloading torchvision (1.3MiB) Downloaded sqlalchemy Downloaded timm Downloading gradio-rangeslider (1.2MiB) Downloading sentencepiece (1.2MiB) Downloaded dashscope Downloaded networkx Downloading kaldifst (1.0MiB) Downloaded cython Downloaded resampy Downloaded numba Downloaded misaki Downloaded jax Downloaded samplerate Downloaded hf-xet Downloaded kaldifst Downloaded sentencepiece Downloaded cryptography Downloaded gradio-rangeslider Downloaded wetext Downloaded torchvision Downloaded pydantic-core Downloaded thinc Downloaded ultralytics Downloaded open-clip-torch Downloaded diffusers Downloaded onnxruntime-gpu Downloaded blis Downloaded pillow Downloaded virtualenv Downloaded pywin32 Downloaded sympy Downloaded scikit-learn Downloaded matplotlib Downloaded babel Downloaded ruff Downloaded espeakng-loader Downloaded pygame Downloaded pandas Downloaded onnx Downloaded onnx-weekly Downloaded numpy Downloaded scikit-image Downloaded transformers Downloaded sherpa-onnx-core Downloaded spacy Downloaded decord Downloaded imageio-ffmpeg Downloaded scipy Downloaded llvmlite Downloaded opencv-python-headless Downloaded opencv-python Downloaded polars-runtime-32 Downloaded gradio Downloaded jaxlib Downloaded taichi Downloaded torch Prepared 272 packages in 27.56s Installed 309 packages in 7.47s
absl-py==2.5.0
accelerate==1.15.0
addict==2.4.0
aiofiles==24.1.0
aiohappyeyeballs==2.7.1
aiohttp==3.14.3
aiosignal==1.4.0
albucore==0.0.24
albumentations==2.0.8
alembic==1.20.0
annotated-doc==0.0.5
annotated-types==0.8.0
antlr4-python3-runtime==4.9.3
anyascii==0.3.3
anyio==4.15.1
apprise==1.12.0
asteroid-filterbanks==0.4.0
attrs==26.1.0
audio-separator==0.36.1
audioread==3.1.0
av==18.1.0
babel==2.18.0
backports-tarfile==1.2.0
beartype==0.18.5
blis==1.3.3
catalogue==2.0.10
certifi==2026.7.22
cffi==2.1.1
cfgv==3.5.0
charset-normalizer==3.5.1
chumpy==0.71 (from https://github.com/deepbeepmeep/chumpy/releases/download/v0.71/chumpy-0.71-py3-none-any.whl)
click==8.5.0
cloudpathlib==0.25.0
cloudpickle==3.1.2
colorama==0.4.6
colorlog==6.12.0
confection==1.3.3
conformer==0.3.2
contourpy==1.3.3
contractions==0.1.73
cryptography==50.0.1
csvw==4.1.0
cycler==0.12.1
cymem==2.0.13
cython==3.3.0
dashscope==1.27.6
decorator==4.4.2
decord==0.6.0
diffq-fixed==0.2.4
diffusers==0.36.0
dill==0.4.1
dlinfo==2.0.0
docopt==0.6.2
easydict==1.13
einops==0.8.2
einshape==1.0
einx==0.4.3
espeakng-loader==0.2.4
facexlib==0.3.0
fastapi==0.141.1
ffmpeg-python==0.2.0
ffmpy==1.0.0
filelock==4.0.1
filterpy==1.4.5
fla-core==0.4.1
flash-linear-attention==0.4.1
flatbuffers==25.12.19
fonttools==4.65.0
frozendict==2.4.7
frozenlist==1.8.0
fsspec==2026.9.0
ftfy==6.3.1
fugashi==1.5.2
future==1.0.0
gguf==0.17.1
gitdb==4.0.12
gitpython==3.1.45
gradio==5.29.0
gradio-client==1.10.0
gradio-rangeslider==0.0.8
greenlet==3.5.6
groovy==0.1.2
h11==0.16.0
hf-xet==1.6.0
httpcore==1.0.9
httpx==0.28.1
httpx-sse==0.4.3
huggingface-hub==0.36.2
hydra-core==1.3.7
hyperpyyaml==1.2.3
identify==2.6.19
idna==3.20
imageio==2.37.4
imageio-ffmpeg==0.6.0
importlib-metadata==9.0.1
importlib-resources==7.1.0
insightface==0.7.3 (from https://github.com/deepbeepmeep/insightface/releases/download/insightface/insightface-0.7.3-cp311-cp311-win_amd64.whl)
iopath==0.1.10
isodate==0.7.2
jaraco-classes==3.4.0
jaraco-context==6.1.2
jaraco-functools==4.6.0
jax==0.10.2
jaxlib==0.10.2
jinja2==3.1.6
joblib==1.6.0
jsonschema==4.26.0
jsonschema-specifications==2025.9.1
julius==0.2.8
kaldifst==1.8.0
keyring==25.7.0
kiwisolver==1.5.1
language-tags==1.3.1
lazy-loader==0.5
librosa==0.11.0
lightning==2.6.6
lightning-utilities==0.15.3
llvmlite==0.49.0
loguru==0.7.3
mako==1.4.1
markdown==3.10.3
markdown-it-py==4.2.0
markupsafe==3.0.3
matplotlib==3.11.2
mcp==1.30.0
mdurl==0.1.2
mido==1.3.3
mir-eval==0.8.2
misaki==0.9.4
ml-collections==1.1.0
ml-dtypes==0.6.0
mmgp==3.8.0
more-itertools==11.1.0
moviepy==1.0.3
mpmath==1.3.0
msgpack==1.2.2
multidict==6.9.0
munch==4.0.0
murmurhash==1.0.15
mutagen==1.48.1
narwhals==2.26.0
networkx==3.6.1
ninja==1.13.2
nodeenv==1.10.0
num2words==0.5.14
numba==0.67.0
numpy==2.1.2
nvidia-ml-py==13.610.43
oauthlib==3.3.1
omegaconf==2.3.1
onnx==1.23.0
onnx-weekly==1.24.0.dev20260914
onnx2torch-py313==1.6.0
onnxruntime-gpu==1.25.0.dev20260210001
open-clip-torch==3.3.0
openai-whisper==20250625
opencv-python==5.0.0.93
opencv-python-headless==5.0.0.93
opt-einsum==3.4.0
optimum-quanto==0.2.7
optuna==5.0.0
orjson==3.11.6
pandas==2.3.3
peft==0.17.0
phonemizer-fork==3.3.2
piexif==1.1.3
pillow==11.3.0
platformdirs==4.11.11
polars==1.44.2
polars-runtime-32==1.44.2
pooch==1.9.0
portalocker==4.3.2
pre-commit==2.3.0
preshed==3.0.13
pretty-midi==0.2.11.post0
prettytable==3.18.0
primepy==1.3
proglog==0.1.12
propcache==0.5.4
protobuf==7.36.2
psutil==7.2.2
pyahocorasick==2.3.1
pyannote-audio==3.3.2
pyannote-core==6.0.1
pyannote-database==6.1.1
pyannote-metrics==3.2.1
pyannote-pipeline==4.0.0
pycocotools==2.0.11
pycparser==3.0
pydantic==2.11.10
pydantic-core==2.33.2
pydantic-settings==2.15.0
pydub==0.25.1
pygame==2.6.1
pygments==2.21.0
pyjwt==2.14.0
pyloudnorm==0.2.0
pymatting==1.1.16
pyparsing==3.3.2
pysmartdl==1.3.4
python-dateutil==2.9.0.post0
python-dotenv==1.2.3
python-multipart==0.0.32
pytorch-lightning==2.6.6
pytorch-metric-learning==2.9.0
pytz==2026.3.post1
pyvers==0.2.3
pywin32==312
pywin32-ctypes==0.2.3
pyyaml==6.0.3
rdflib==7.6.0
referencing==0.37.0
regex==2026.9.10
rembg==2.0.65
requests==2.34.2
requests-oauthlib==2.0.0
resampy==0.4.3
rfc3986==1.5.0
rich==15.0.0
rotary-embedding-torch==0.6.5
rpds-py==2026.6.3
ruamel-yaml==0.18.17
ruamel-yaml-clib==0.2.15
ruff==0.16.8
s3tokenizer==0.3.0
safehttpx==0.1.7
safetensors==0.8.0
samplerate==0.1.0
scikit-image==0.26.0
scikit-learn==1.9.1
scipy==1.17.1
segment-anything==1.0
segments==2.4.0
semantic-version==2.10.0
semver==3.1.0
sentencepiece==0.2.2
setuptools==84.0.0
shellingham==1.5.4
sherpa-onnx==1.13.2
sherpa-onnx-core==1.13.2
simsimd==6.5.16
six==1.17.0
smart-open==8.0.1
smmap==5.0.3
smplfitter==0.2.10 (from https://github.com/deepbeepmeep/smplfitter/releases/download/v0.2.10/smplfitter-0.2.10-py3-none-any.whl)
sortedcontainers==2.4.0
sounddevice==0.5.6
soundfile==0.14.0
soxr==1.1.0
spacy==3.8.16
spacy-legacy==3.0.12
spacy-loggers==1.0.5
spacy-pkuseg==1.0.1
speechbrain==1.0.3
sqlalchemy==2.0.54
srsly==2.5.3
sse-starlette==3.4.11
starlette==0.52.1
stringzilla==4.0.14
sympy==1.14.0
symusic==0.6.0
tabulate==0.10.0
taichi==1.7.4
tensorboardx==2.6.5
tensordict==0.14.2
termcolor==3.3.0
textsearch==0.0.24
thinc==8.3.13
threadpoolctl==3.7.0
tifffile==2026.3.3
tiktoken==0.14.0
timm==1.0.29
tokenizers==0.21.4
toml==0.10.2
tomlkit==0.13.3
torch==2.14.0
torch-audiomentations==0.12.0
torch-pitch-shift==1.2.5
torchaudio==2.11.0
torchcodec==0.10.0
torchdiffeq==0.2.5
torchmetrics==1.9.0
torchvision==0.29.0
tqdm==4.70.1
transformers==4.54.0
typer==0.27.2
typing-extensions==4.16.0
typing-inspection==0.4.4
tzdata==2026.4
ultralytics==8.4.155
ultralytics-platform==0.1.48
ultralytics-thop==2.1.6
unidic-lite==1.0.8
uritemplate==4.2.0
urllib3==2.8.0
uvicorn==0.53.0
vector-quantize-pytorch==1.27.19
virtualenv==16.7.12
wasabi==1.1.3
wcwidth==0.8.4
weasel==1.0.0
websocket-client==1.9.2
websockets==15.0.1
wetext==0.1.2
win32-setctime==1.2.0
wrapt==2.4.1
xxhash==4.0.1
yarl==1.25.1
zipp==4.1.0 Using Python 3.11.14 environment at: venv Resolved 4 packages in 191ms Downloading xformers (103.4MiB) Downloading torch (3.0GiB) Downloading torchaudio (4.4MiB) Downloading torchvision (7.3MiB) Downloaded torchaudio Downloaded torchvision Downloaded xformers Downloaded torch Prepared 4 packages in 57.46s Uninstalled 3 packages in 720ms Installed 4 packages in 1.26s

torch==2.14.0

torch==2.7.1+cu128

torchaudio==2.11.0

torchaudio==2.7.1+cu128

torchvision==0.29.0

torchvision==0.22.1+cu128
xformers==0.0.30 Using Python 3.11.14 environment at: venv Resolved 2 packages in 105ms Prepared 1 package in 187ms Uninstalled 1 package in 25ms Installed 1 package in 31ms

setuptools==84.0.0

setuptools==69.5.1 Using Python 3.11.14 environment at: venv Resolved 2 packages in 114ms Downloading triton-windows (40.0MiB) Downloaded triton-windows Prepared 1 package in 840ms Installed 1 package in 33ms
triton-windows==3.3.1.post19 Using Python 3.11.14 environment at: venv × No solution found when resolving dependencies: ╰─▶ Because only sageattention==2.2.0+cu128torch2.7.1 is available and sageattention==2.2.0+cu128torch2.7.1 has no wheels with a matching Python version tag (e.g., cp311), we can conclude that all versions of sageattention cannot be used. And because you require sageattention, we can conclude that your requirements are unsatisfiable. Could not install Wan2GP (StabilityMatrix.Core.Exceptions.ProcessException: pip install failed with code 1: 'Using Python 3.11.14 environment at: venv\n \u00d7 No solution found when resolving dependencies:\n \u2570\u2500\u25b6 Because only sageattention==2.2.0+cu128torch2.7.1 is available and\n sageattention==2.2.0+cu128torch2.7.1 has no wheels with a matching\n Python version tag (e.g., cp311), we can conclude that all versions of\n sageattention cannot be used.\n And because you require sageattention, we can conclude that your\n requirements are unsatisfiable.\n' at StabilityMatrix.Core.Python.UvVenvRunner.PipInstall(ProcessArgs args, Action1 outputDataReceived) at StabilityMatrix.Core.Models.Packages.Wan2GP.InstallNvidiaAsync(IPyVenvRunner venvRunner, IProgress1 progress, Action1 onConsoleOutput) at StabilityMatrix.Core.Models.Packages.Wan2GP.InstallPackage(String installLocation, InstalledPackage installedPackage, InstallPackageOptions options, IProgress1 progress, Action1 onConsoleOutput, CancellationToken cancellationToken) at StabilityMatrix.Core.Models.Packages.Wan2GP.InstallPackage(String installLocation, InstalledPackage installedPackage, InstallPackageOptions options, IProgress1 progress, Action1 onConsoleOutput, CancellationToken cancellationToken) at StabilityMatrix.Core.Models.PackageModification.InstallPackageStep.ExecuteAsync(IProgress1

Version

2.16.4

What Operating System are you using?

Windows

Source: LykosAI/StabilityMatrix

View original on GitHub View discussion on GitHub