Sygil/Sygil-Diffusion

Instalación básica con Diffusers:

pip install -U diffusers transformers accelerate

Uso con DiffusionPipeline:

import torch
from diffusers import DiffusionPipeline

# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("Sygil/Sygil-Diffusion", dtype=torch.bfloat16, device_map="cuda")

prompt = "environment art, realistic"
image = pipe(prompt).images[0]

Ejemplo recomendado con StableDiffusionPipeline y DPMSolverMultistepScheduler:

pip install diffusers transformers accelerate scipy safetensors

import torch
from diffusers import StableDiffusionPipeline, DPMSolverMultistepScheduler

model_id = "Sygil/Sygil-Diffusion"

# Use the DPMSolverMultistepScheduler (DPM-Solver++) scheduler here instead
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16)
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
pipe = pipe.to("cuda")

prompt = "a beautiful illustration of a fantasy forest"
image = pipe(prompt).images[0]
image.save("fantasy_forest_illustration.png")

Para mejor rendimiento, el autor recomienda instalar xformers. En GPUs con poca VRAM, se puede activar pipe.enable_attention_slicing() después de mover el pipeline a CUDA, reduciendo memoria a cambio de velocidad.