Transformers fix

[Artheriax]

Summary

Fix AttributeError: 'CLIPTextModel' object has no attribute 'text_model' caused by
transformers >= 5.9 restructuring CLIPTextModel internals.

In newer transformers releases, embeddings, encoder, and final_layer_norm moved
from a nested text_model submodule directly onto CLIPTextModel.
CLIPTextModelWithProjection is unaffected — it still wraps an inner CLIPTextModel
via .text_model.

9 files changed (9 edits):

• modules/modelSetup/BaseStableDiffusionSetup.py — .text_model.embeddings → .embeddings
• modules/modelSetup/BaseStableDiffusionXLSetup.py — .text_model.embeddings → .embeddings (te1 only)
• modules/modelSetup/BaseFluxSetup.py — .text_model.embeddings → .embeddings (te1 only)
• modules/modelSetup/BaseHunyuanVideoSetup.py — .text_model.embeddings → .embeddings (te2 only)
• modules/modelSetup/BaseWuerstchenSetup.py — .text_model.embeddings → .embeddings (prior)
• modules/modelLoader/stableDiffusionXL/StableDiffusionXLModelLoader.py — .text_model.embeddings → .embeddings (ckpt path)
• modules/model/WuerstchenModel.py — .text_model.final_layer_norm → .final_layer_norm
• modules/modelSampler/WuerstchenSampler.py — .text_model.final_layer_norm → .final_layer_norm
• modules/model/util/clip_util.py — added hasattr(text_encoder, 'text_model') guard

Test plan

• pre-commit run --all-files passes
• Launched the affected UI or script and exercised the change
• Tested with at least one real preset / config when relevant (note which: SDXL lora)

AI assistance

• AI-assisted to find the cause of the bug.

[dxqb]

if you use the preview branch, please mention it. The preview branch is known to be broken for SDXL or any model that uses CLIP.
This fix is not sufficient. The entire scope you can find here: #1506