Trying to get this working under Windows

[SoftologyPro]

I install without Apex using this batch file...

@echo off

cd
echo *** Deleting Lumina-Video directory if it exists
if exist Lumina-Video\. rd /S /Q Lumina-Video

echo *** Cloning Lumina-Video repository
git clone https://github.com/Alpha-VLLM/Lumina-Video
cd Lumina-Video

echo *** Creating venv
python -m venv venv
call venv\scripts\activate.bat

echo *** Installing requirments
python -m pip install --upgrade pip
pip install -r requirements.txt
pip install https://huggingface.co/datasets/Softology-Pro/VoC/resolve/main/flash_attn-2.7.1.post1+cu124torch2.5.1cxx11abiFALSE-cp310-cp310-win_amd64.whl
pip install imageio
pip install https://huggingface.co/datasets/Softology-Pro/VoC/resolve/main/triton-3.0.0-cp310-cp310-win_amd64.whl

echo *** Installing GPU Torch
pip uninstall -y torch
pip uninstall -y torch
pip install --no-cache-dir --ignore-installed --force-reinstall --no-warn-conflicts torch==2.5.1+cu124 torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124

rem echo *** Installing Apex
rem git clone https://github.com/NVIDIA/apex
rem cd apex
rem pip install wheel
rem pip install setuptools
rem pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" ./

echo *** Downloading models
if not exist "ckpts" mkdir "ckpts"
if not exist "ckpts\f24R960" mkdir "ckpts\f24R960"
huggingface-cli download --resume-download Alpha-VLLM/Lumina-Video-f24R960 --local-dir ./ckpts/f24R960

call venv\scripts\deactivate.bat
cd ..

echo *** Finished Lumina-Video install
echo.
echo *** Scroll up and check for errors.  Do not assume it worked.
pause

And then run with this batch file...

@echo off
set USE_LIBUV=0
set TORCHDYNAMO_VERBOSE=1 
set PL_TORCH_DISTRIBUTED_BACKEND=gloo
cd Lumina-Video
call venv\scripts\activate.bat
python -u generate.py --ckpt ./ckpts/f24R960 --resolution 1248x704 --fps 24 --frames 96 --prompt "Shrek eating pizza" --neg_prompt "" --sample_config f24F96R960
call venv\scripts\deactivate.bat

But it gives these errors...

D:\Tests\Lumina-Video>run
D:\Tests\Lumina-Video\Lumina-Video\models\components.py:6: UserWarning: Cannot import apex RMSNorm, switch to vanilla implementation
  warnings.warn("Cannot import apex RMSNorm, switch to vanilla implementation")
D:\Tests\Lumina-Video\Lumina-Video\venv\lib\site-packages\torch\distributed\distributed_c10d.py:730: UserWarning: Attempted to get default timeout for nccl backend, but NCCL support is not compiled
  warnings.warn(
[W312 20:07:28.000000000 socket.cpp:752] [c10d] The client socket has failed to connect to [kubernetes.docker.internal]:10000 (system error: 10049 - The requested address is not valid in its context.).
Traceback (most recent call last):
  File "D:\Tests\Lumina-Video\Lumina-Video\generate.py", line 238, in <module>
    main()
  File "D:\Tests\Lumina-Video\Lumina-Video\generate.py", line 224, in main
    dist.init_process_group("nccl")
  File "D:\Tests\Lumina-Video\Lumina-Video\venv\lib\site-packages\torch\distributed\c10d_logger.py", line 83, in wrapper
    return func(*args, **kwargs)
  File "D:\Tests\Lumina-Video\Lumina-Video\venv\lib\site-packages\torch\distributed\c10d_logger.py", line 97, in wrapper
    func_return = func(*args, **kwargs)
  File "D:\Tests\Lumina-Video\Lumina-Video\venv\lib\site-packages\torch\distributed\distributed_c10d.py", line 1527, in init_process_group
    default_pg, _ = _new_process_group_helper(
  File "D:\Tests\Lumina-Video\Lumina-Video\venv\lib\site-packages\torch\distributed\distributed_c10d.py", line 1750, in _new_process_group_helper
    raise RuntimeError("Distributed package doesn't have NCCL built in")
RuntimeError: Distributed package doesn't have NCCL built in
D:\Tests\Lumina-Video\Lumina-Video>

Any ideas how I can get this working under Windows (on a 24GB 4090)?