import gradio as gr from pipeline_rf import RectifiedFlowPipeline import torch from torchvision.transforms import Compose, Resize, CenterCrop, ToTensor, Normalize import torch.nn.functional as F from diffusers import StableDiffusionXLImg2ImgPipeline import time import copy import numpy as np def merge_dW_to_unet(pipe, dW_dict, alpha=1.0): _tmp_sd = pipe.unet.state_dict() for key in dW_dict.keys(): _tmp_sd[key] += dW_dict[key] * alpha pipe.unet.load_state_dict(_tmp_sd, strict=False) return pipe def get_dW_and_merge(pipe_rf, lora_path='Lykon/dreamshaper-7', save_dW = False, base_sd='runwayml/stable-diffusion-v1-5', alpha=1.0): # get weights of base sd models from diffusers import DiffusionPipeline _pipe = DiffusionPipeline.from_pretrained( base_sd, torch_dtype=torch.float16, safety_checker = None, ) sd_state_dict = _pipe.unet.state_dict() # get weights of the customized sd models, e.g., the aniverse downloaded from civitai.com _pipe = DiffusionPipeline.from_pretrained( lora_path, torch_dtype=torch.float16, safety_checker = None, ) lora_unet_checkpoint = _pipe.unet.state_dict() # get the dW dW_dict = {} for key in lora_unet_checkpoint.keys(): dW_dict[key] = lora_unet_checkpoint[key] - sd_state_dict[key] # return and save dW dict if save_dW: save_name = lora_path.split('/')[-1] + '_dW.pt' torch.save(dW_dict, save_name) pipe_rf = merge_dW_to_unet(pipe_rf, dW_dict=dW_dict, alpha=alpha) pipe_rf.vae = _pipe.vae pipe_rf.text_encoder = _pipe.text_encoder return dW_dict pipe = StableDiffusionXLImg2ImgPipeline.from_pretrained( "stabilityai/stable-diffusion-xl-refiner-1.0", torch_dtype=torch.float16, variant="fp16", use_safetensors=True ) pipe = pipe.to("cuda") insta_pipe = RectifiedFlowPipeline.from_pretrained("XCLiu/instaflow_0_9B_from_sd_1_5", torch_dtype=torch.float16) dW_dict = get_dW_and_merge(insta_pipe, lora_path="Lykon/dreamshaper-7", save_dW=False, alpha=1.0) insta_pipe.to("cuda") global img @torch.no_grad() def set_new_latent_and_generate_new_image(seed, prompt, randomize_seed, num_inference_steps=1, guidance_scale=0.0): print('Generate with input seed') global img negative_prompt="" if randomize_seed: seed = np.random.randint(0, 2**32) seed = int(seed) num_inference_steps = int(num_inference_steps) guidance_scale = float(guidance_scale) print(seed, num_inference_steps, guidance_scale) t_s = time.time() generator = torch.manual_seed(seed) images = insta_pipe(prompt=prompt, num_inference_steps=1, guidance_scale=0.0, generator=generator).images inf_time = time.time() - t_s img = copy.copy(np.array(images[0])) return images[0], inf_time, seed @torch.no_grad() def refine_image_512(prompt): print('Refine with SDXL-Refiner (512)') global img t_s = time.time() img = torch.tensor(img).unsqueeze(0).permute(0, 3, 1, 2) / 255.0 img = img.permute(0, 2, 3, 1).squeeze(0).cpu().numpy() new_image = pipe(prompt, image=img).images[0] print('time consumption:', time.time() - t_s) new_image = np.array(new_image) * 1.0 / 255. img = copy.copy(new_image) return new_image with gr.Blocks() as gradio_gui: gr.Markdown( """ # InstaFlow! One-Step Stable Diffusion with Rectified Flow [[paper]](https://arxiv.org/abs/2309.06380) ## This is a demo of one-step InstaFlow-0.9B with [dreamshaper-7](https://huggingface.co/Lykon/dreamshaper-7) (a LoRA that improves image quality) and measures the inference time. """) with gr.Row(): with gr.Column(scale=0.4): with gr.Group(): gr.Markdown("Generation from InstaFlow-0.9B") im = gr.Image() with gr.Column(scale=0.4): inference_time_output = gr.Textbox(value='0.0', label='Inference Time with One-Step InstaFlow (Second)') seed_input = gr.Textbox(value='101098274', label="Random Seed") randomize_seed = gr.Checkbox(label="Randomly Sample a Random Seed", value=True) prompt_input = gr.Textbox(value='A high-resolution photograph of a waterfall in autumn; muted tone', label="Prompt") new_image_button = gr.Button(value="One-Step Generation with InstaFlow and the Random Seed") new_image_button.click(set_new_latent_and_generate_new_image, inputs=[seed_input, prompt_input, randomize_seed], outputs=[im, inference_time_output, seed_input]) refine_button_512 = gr.Button(value="Refine One-Step Generation with SDXL Refiner (Resolution: 512)") refine_button_512.click(refine_image_512, inputs=[prompt_input], outputs=[im]) gradio_gui.launch()