Updated Installers Just Published - Torch 2.8, CUDA 12.9 Supports every GPU including RTX 5000 series with xFormers - Windows, RunPod & Massed Compute (Linux)

[FurkanGozukara]

I just updated my all installers for Windows, RunPod and Massed Compute

Linux users can use Massed Compute scripts

Download from here : https://www.patreon.com/posts/86307255

Only installation is upgraded rest is same as this repo original Automatic1111 Web UI no changes

28 March 2028 V12

• Installers completey modernized
• Now entire installation is made with uv installer on Windows, RunPod and Massed Compute (Linux users)
• Thus it is like 100x faster now
• Everything upgraded to Torch 2.8 and CUDA 12.9 with pre-compiled wheels
• ControlNet model downloader issues fixed
• Installation, ControlNet and Adetailer tested on Windows and RunPod and Massed Compute (Linux) and all working perfect
• This is the original official latest version Automatic1111 without new features or changes
• Make a fresh install
• ControlNet downloader is downloading 53 different SD 1.5 and SDXL ControlNet models automatically for you
• Installer automatically installs working ControlNet latest and Adetailer

[megvadulthangya]

It’s interesting to see these third-party installers, but since the project is AGPL-3.0, are the scripts for these 'modernized installers' also publicly available for the community to audit and improve? It’s important that we keep the open-source spirit alive and don't gate-keep basic setup steps behind paywalls.

[Ineman]

Is it possible to use xformers with blackwell? Which version of xformers did you use? When i try to use torch 2.8.0+cu129 and xformers 0.0.32.post2 in python 3.10.19 linux, it makes error.

CUDA error (/__w/xformers/xformers/third_party/flash-attention/hopper/flash_fwd_launch_template.h:188): invalid argument
Exception in thread MemMon:
Traceback (most recent call last):
  File "/venv/main/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
    self.run()
  File "/workspace/stable-diffusion-webui/modules/memmon.py", line 53, in run
    free, total = self.cuda_mem_get_info()
  File "/workspace/stable-diffusion-webui/modules/memmon.py", line 33, in cuda_mem_get_info
    index = self.device.index if self.device.index is not None else torch.cuda.current_device()
  File "/venv/main/lib/python3.10/site-packages/torch/cuda/__init__.py", line 1072, in current_device
    return torch._C._cuda_getDevice()
torch.AcceleratorError: CUDA error: driver shutting down
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.