used to run in a 5090

[Nicolas-Melluso]

Description

This PR restores compatibility with the legacy Stable Diffusion repository layout required by some setups (including my RTX 5090 environment), and prevents startup/import issues caused by missing taming-transformers path registration.

• a simple description of what you're trying to accomplish

Switched default STABLE_DIFFUSION_REPO from Stability-AI/stablediffusion back to CompVis/stable-diffusion in modules/launch_utils.py.

Updated default STABLE_DIFFUSION_COMMIT_HASH to 21f890f9da3cfbeaba8e2ac3c425ee9e998d5229 in modules/launch_utils.py.

Added taming-transformers path registration in modules/paths.py to ensure taming imports are found at startup.

Made OpenCLIP type check more defensive in modules/sd_hijack.py using getattr(..., "FrozenOpenCLIPEmbedder", type(None)) to avoid hard failures when symbol resolution differs.

Updated local webui-user.bat defaults for my environment (PYTHON, --api, SD repo/hash) for reproducible launch behavior.

• a summary of changes in code
• which issues it fixes, if any

Fixes startup/import errors in environments expecting the CompVis + taming-transformers structure.

Fixes a fragile class comparison path in OpenCLIP hijack logic.

Screenshots/videos:

N/A

Checklist:

• I have read contributing wiki page
• I have performed a self-review of my own code
• My code follows the style guidelines
• My code passes tests