import os import gradio as gr from gradio_imageslider import ImageSlider import argparse from SUPIR.util import HWC3, upscale_image, fix_resize, convert_dtype import numpy as np import torch from SUPIR.util import create_SUPIR_model, load_QF_ckpt from PIL import Image from llava.llava_agent import LLavaAgent from CKPT_PTH import LLAVA_MODEL_PATH import einops import copy import time import spaces from huggingface_hub import hf_hub_download from diffusers import AnimateDiffPipeline, DDIMScheduler, MotionAdapter from diffusers.utils import export_to_gif from diffusers.utils import export_to_video from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline import uuid hf_hub_download(repo_id="laion/CLIP-ViT-bigG-14-laion2B-39B-b160k", filename="open_clip_pytorch_model.bin", local_dir="laion_CLIP-ViT-bigG-14-laion2B-39B-b160k") hf_hub_download(repo_id="camenduru/SUPIR", filename="sd_xl_base_1.0_0.9vae.safetensors", local_dir="yushan777_SUPIR") hf_hub_download(repo_id="camenduru/SUPIR", filename="SUPIR-v0F.ckpt", local_dir="yushan777_SUPIR") hf_hub_download(repo_id="camenduru/SUPIR", filename="SUPIR-v0Q.ckpt", local_dir="yushan777_SUPIR") hf_hub_download(repo_id="RunDiffusion/Juggernaut-XL-Lightning", filename="Juggernaut_RunDiffusionPhoto2_Lightning_4Steps.safetensors", local_dir="RunDiffusion_Juggernaut-XL-Lightning") parser = argparse.ArgumentParser() parser.add_argument("--opt", type=str, default='options/SUPIR_v0.yaml') parser.add_argument("--ip", type=str, default='127.0.0.1') parser.add_argument("--port", type=int, default='6688') parser.add_argument("--no_llava", action='store_true', default=False) parser.add_argument("--use_image_slider", action='store_true', default=False) parser.add_argument("--log_history", action='store_true', default=False) parser.add_argument("--loading_half_params", action='store_true', default=False) parser.add_argument("--use_tile_vae", action='store_true', default=False) parser.add_argument("--encoder_tile_size", type=int, default=512) parser.add_argument("--decoder_tile_size", type=int, default=64) parser.add_argument("--load_8bit_llava", action='store_true', default=False) args = parser.parse_args() server_ip = args.ip server_port = args.port use_llava = not args.no_llava if torch.cuda.device_count() > 0: if torch.cuda.device_count() >= 2: SUPIR_device = 'cuda:0' LLaVA_device = 'cuda:1' elif torch.cuda.device_count() == 1: SUPIR_device = 'cuda:0' LLaVA_device = 'cuda:0' else: SUPIR_device = 'cpu' LLaVA_device = 'cpu' # load SUPIR model, default_setting = create_SUPIR_model(args.opt, SUPIR_sign='Q', load_default_setting=True) if args.loading_half_params: model = model.half() if args.use_tile_vae: model.init_tile_vae(encoder_tile_size=args.encoder_tile_size, decoder_tile_size=args.decoder_tile_size) model = model.to(SUPIR_device) model.first_stage_model.denoise_encoder_s1 = copy.deepcopy(model.first_stage_model.denoise_encoder) model.current_model = 'v0-Q' ckpt_Q, ckpt_F = load_QF_ckpt(args.opt) # load LLaVA #if use_llava: #llava_agent = LLavaAgent(LLAVA_MODEL_PATH, device=LLaVA_device, load_8bit=args.load_8bit_llava, load_4bit=False) #else: #llava_agent = None # Available adapters (replace with your actual adapter names) adapter_options = { "zoom-out":"guoyww/animatediff-motion-lora-zoom-out", "zoom-in":"guoyww/animatediff-motion-lora-zoom-in", "pan-left":"guoyww/animatediff-motion-lora-pan-left", "pan-right":"guoyww/animatediff-motion-lora-pan-right", "roll-clockwise":"guoyww/animatediff-motion-lora-rolling-clockwise", "roll-anticlockwise":"guoyww/animatediff-motion-lora-rolling-anticlockwise", "tilt-up":"guoyww/animatediff-motion-lora-tilt-up", "tilt-down":"guoyww/animatediff-motion-lora-tilt-down" } def load_cached_examples(): examples = [ ["a cat playing with a ball of yarn", "blurry", 7.5, 12, ["zoom-in"]], ["a dog running in a field", "dark, indoors", 8.0, 8, ["pan-left", "tilt-up"]], ] return examples device = "cuda" adapter = MotionAdapter.from_pretrained("guoyww/animatediff-motion-adapter-v1-5-2", torch_dtype=torch.float16) model_id = "SG161222/Realistic_Vision_V5.1_noVAE" pipe = AnimateDiffPipeline.from_pretrained(model_id, motion_adapter=adapter, torch_dtype=torch.float16).to(device) scheduler = DDIMScheduler.from_pretrained( model_id, subfolder="scheduler", clip_sample=False, timestep_spacing="linspace", beta_schedule="linear", steps_offset=1, ) pipe.scheduler = scheduler @spaces.GPU def generate_video(prompt,negative_prompt, guidance_scale, num_inference_steps, adapter_choices): pipe.to(device) # Set adapters based on user selection if adapter_choices: for i in range(len(adapter_choices)): adapter_name = adapter_choices[i] pipe.load_lora_weights( adapter_options[adapter_name], adapter_name=adapter_name, ) pipe.set_adapters(adapter_choices, adapter_weights=[1.0] * len(adapter_choices)) print(adapter_choices) output = pipe( prompt=prompt, negative_prompt=negative_prompt, num_frames=16, guidance_scale=guidance_scale, num_inference_steps=num_inference_steps, ) name = str(uuid.uuid4()).replace("-", "") path = f"/tmp/{name}.mp4" export_to_video(output.frames[0], path, fps=10) return path iface = gr.Interface( theme=gr.themes.Soft(primary_hue="cyan", secondary_hue="teal"), fn=generate_video, inputs=[ gr.Textbox(label="Prompt"), gr.Textbox(label="Negative Prompt"), gr.Slider(minimum=0.5, maximum=10, value=7.5, label="Guidance Scale"), gr.Slider(minimum=4, maximum=24, step=4, value=4, label="Inference Steps"), gr.CheckboxGroup(adapter_options.keys(), label="Adapter Choice",type='value'), ], outputs=gr.Video(label="Generated Video"), examples = [ ["Urban ambiance, man walking, neon lights, rain, wet floor, high quality", "bad quality", 7.5, 24, []], ["Nature, farms, mountains in background, drone shot, high quality","bad quality" ,8.0, 24, []], ], cache_examples=True ) iface.launch()