Describe the bug

Anima models like the turbo and aesthetic models, as well as community finetunes like real dream use the model.diffusion_model prefix, but the CosmosTransformer3DModel.from_single_file() method fails to detect this prefix, and hence it fails to load the file.

I will be putting up a PR for this very soon, please guide me if my PR needs improvement.

Reproduction

from diffusers import CosmosTransformer3DModel
transformer = CosmosTransformer3DModel.from_single_file(
    "https://huggingface.co/circlestone-labs/Anima/blob/main/split_files/diffusion_models/anima-turbo-v1.1.safetensors"
)

Logs

System Info

• 🤗 Diffusers version: 0.41.0.dev0
• Platform: Windows-10-10.0.26200-SP0
• Running on Google Colab?: No
• Python version: 3.11.9
• PyTorch version (GPU?): 2.10.0+cu130 (True)
• Huggingface_hub version: 1.33.0
• Transformers version: 5.5.4
• Accelerate version: 1.6.0
• PEFT version: 0.19.0
• gguf version: 0.17.1
• optimum-quanto version: 0.2.7
• Safetensors version: 0.8.0
• xFormers version: not installed
• Accelerator: NVIDIA GeForce RTX 4090, 24564 MiB
• Using GPU in script?: No
• Using distributed or parallel set-up in script?: No

Who can help?

@DN6