Detects SDXL checkpoints (e.g. stabilityai/sdxl-turbo) from the repo's
file listing and handles their dual CLIP text encoder + pooled
micro-conditioning across ESD, MACE, CA, CoGFD, SSD, TraSCE. Also fixes
breakage from transformers >=4.45 flattening CLIPTextModel (removed
`.text_model` submodule) in AdvUnlearn, and updates MMA-Diffusion/P4D
for current diffusers/transformers APIs.
Excludes local-only third-party clones (LLaVA-NeXT, diffusion-mu-attack)
and the eval-unlearn package's own nested repo from tracking, so they
aren't accidentally swept into this repo by a future `git add -A`.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
