inference_flux.py

import os
from pipelines.pipeline_flux import FluxMultiDiffusionPipeline
import diffusers
import torch
from diffusers.utils import load_image
from PIL import Image
Image.MAX_IMAGE_PIXELS = None
import numpy as np
import cv2
from pathlib import Path

base_model = "black-forest-labs/FLUX.1-dev"

inference_dtype = torch.bfloat16

base = FluxMultiDiffusionPipeline.from_pretrained(
    base_model,
    torch_dtype=inference_dtype,
)

base.enable_model_cpu_offload(0)

instanceprompt = ("A breathtaking, ultra-wide panoramic landscape of jagged obsidian mountain peaks at twilight. The valleys are filled with flowing rivers of glowing violet bioluminescent mist. Above, a sprawling cosmic sky features a massive, detailed nebula in shades of deep teal and gold, with scattered crystalline stars. Photorealistic, cinematic lighting, 8k resolution, highly detailed rock textures, atmospheric depth, ethereal glow, wide-angle lens.")

#Path + name 
output_path = Path("./outputs")
output_path.mkdir(parents=True, exist_ok=True)
img_name = "res_flux1"

TARGET_HEIGHT = 2048
TARGET_WIDTH = 2048

image = base(
    prompt = instanceprompt,
    height = 1024,
    width = 1024,
    total_height = TARGET_HEIGHT,
    total_width = TARGET_WIDTH,
    num_inference_steps = 28,
    guidance_scale = 4.0,        #helps FLUX.1-dev follow the prompt
    view_batch_size = 1,
    stride_start= 512,
    stride_end= 512 + 256,
    return_dict = False
)[0][0]

image.save(output_path / f"{img_name}.png")