Vargol/ProteusV0.4

This a an fp16 variant of Proteus V0.4 https://huggingface.co/dataautogpt3/ProteusV0.4 currently under the gpl-v3 licence. i Made by simply loading and sving.

import torch
from diffusers import DiffusionPipeline

pipeline = DiffusionPipeline.from_pretrained("/Volumes/SSD2TB/AI/caches/invoke_models/sdxl/main/ProteusV0.4" , torch_dtype=torch.float16)
pipeline.save_pretrained('ProteusV0.4', safe_serialization=True, variant='fp16')

Use like anyother fp16 variant

pipe = DiffusionPipeline.from_pretrained(
      'Vargol/ProteusV0,4',
      torch_dtype=torch.float16,
      use_safetensors=True,
      variant="fp16",
      ).to('cuda')