Upload app (8).py

app (8).py

@@ -0,0 +1,148 @@
+import spaces
+import gradio as gr
+import os
+import sys
+import argparse
+import random
+import time
+from omegaconf import OmegaConf
+import torch
+import torchvision
+from pytorch_lightning import seed_everything
+from huggingface_hub import hf_hub_download
+from einops import repeat
+import torchvision.transforms as transforms
+from utils.utils import instantiate_from_config
+sys.path.insert(0, "scripts/evaluation")
+from funcs import (
+    batch_ddim_sampling,
+    load_model_checkpoint,
+    get_latent_z,
+    save_videos
+)
+from transformers import pipeline
+
+def download_model():
+    REPO_ID = 'Doubiiu/DynamiCrafter_1024'
+    filename_list = ['model.ckpt']
+    if not os.path.exists('./checkpoints/dynamicrafter_1024_v1/'):
+        os.makedirs('./checkpoints/dynamicrafter_1024_v1/')
+    for filename in filename_list:
+        local_file = os.path.join('./checkpoints/dynamicrafter_1024_v1/', filename)
+        if not os.path.exists(local_file):
+            hf_hub_download(repo_id=REPO_ID, filename=filename, local_dir='./checkpoints/dynamicrafter_1024_v1/', force_download=True)
+
+download_model()
+ckpt_path='checkpoints/dynamicrafter_1024_v1/model.ckpt'
+config_file='configs/inference_1024_v1.0.yaml'
+config = OmegaConf.load(config_file)
+model_config = config.pop("model", OmegaConf.create())
+model_config['params']['unet_config']['params']['use_checkpoint']=False   
+model = instantiate_from_config(model_config)
+assert os.path.exists(ckpt_path), "Error: checkpoint Not Found!"
+model = load_model_checkpoint(model, ckpt_path)
+model.eval()
+model = model.cuda()
+
+# 번역 모델 초기화
+translator = pipeline("translation", model="Helsinki-NLP/opus-mt-ko-en")
+
+@spaces.GPU(duration=300)
+def infer(image, prompt, steps=50, cfg_scale=7.5, eta=1.0, fs=3, seed=123, video_length=2):
+    # 한글 입력 감지 및 번역
+    if any('\u3131' <= char <= '\u318E' or '\uAC00' <= char <= '\uD7A3' for char in prompt):
+        translated = translator(prompt, max_length=512)[0]['translation_text']
+        prompt = translated
+        print(f"Translated prompt: {prompt}")
+
+    resolution = (576, 1024)
+    save_fps = 8
+    seed_everything(seed)
+    transform = transforms.Compose([
+        transforms.Resize(min(resolution)),
+        transforms.CenterCrop(resolution),
+        ])
+    torch.cuda.empty_cache()
+    print('start:', prompt, time.strftime('%Y-%m-%d %H:%M:%S',time.localtime(time.time())))
+    start = time.time()
+    if steps > 60:
+        steps = 60 
+
+    batch_size=1
+    channels = model.model.diffusion_model.out_channels
+    frames = int(video_length * save_fps)  # 비디오 길이에 따른 프레임 수 계산
+    h, w = resolution[0] // 8, resolution[1] // 8
+    noise_shape = [batch_size, channels, frames, h, w]
+
+    # text cond
+    with torch.no_grad(), torch.cuda.amp.autocast():
+        text_emb = model.get_learned_conditioning([prompt])
+    
+        # img cond
+        img_tensor = torch.from_numpy(image).permute(2, 0, 1).float().to(model.device)
+        img_tensor = (img_tensor / 255. - 0.5) * 2
+    
+        image_tensor_resized = transform(img_tensor) #3,256,256
+        videos = image_tensor_resized.unsqueeze(0) # bchw
+        
+        z = get_latent_z(model, videos.unsqueeze(2)) #bc,1,hw
+        
+        img_tensor_repeat = repeat(z, 'b c t h w -> b c (repeat t) h w', repeat=frames)
+    
+        cond_images = model.embedder(img_tensor.unsqueeze(0)) ## blc
+        img_emb = model.image_proj_model(cond_images)
+    
+        imtext_cond = torch.cat([text_emb, img_emb], dim=1)
+    
+        fs = torch.tensor([fs], dtype=torch.long, device=model.device)
+        cond = {"c_crossattn": [imtext_cond], "fs": fs, "c_concat": [img_tensor_repeat]}
+        
+        ## inference
+        batch_samples = batch_ddim_sampling(model, cond, noise_shape, n_samples=1, ddim_steps=steps, ddim_eta=eta, cfg_scale=cfg_scale)
+        ## b,samples,c,t,h,w
+    
+        video_path = './output.mp4'
+        save_videos(batch_samples, './', filenames=['output'], fps=save_fps)
+    return video_path
+
+i2v_examples = [
+    ['prompts/1024/astronaut04.png', 'a man in an astronaut suit playing a guitar', 30, 7.5, 1.0, 6, 123, 2],
+]
+
+css = """#input_img {max-width: 1024px !important} #output_vid {max-width: 1024px; max-height: 576px}"""
+
+with gr.Blocks(analytics_enabled=False, css=css) as dynamicrafter_iface:
+    gr.Markdown("이미지로 영상 생성 테스트 (한글 프롬프트 지원)")
+    with gr.Tab(label='ImageAnimation_576x1024'):
+        with gr.Column():
+            with gr.Row():
+                with gr.Column():
+                    with gr.Row():
+                        i2v_input_image = gr.Image(label="Input Image",elem_id="input_img")
+                    with gr.Row():
+                        i2v_input_text = gr.Text(label='Prompts')
+                    with gr.Row():
+                        i2v_seed = gr.Slider(label='Random Seed', minimum=0, maximum=10000, step=1, value=123)
+                        i2v_eta = gr.Slider(minimum=0.0, maximum=1.0, step=0.1, label='ETA', value=1.0, elem_id="i2v_eta")
+                        i2v_cfg_scale = gr.Slider(minimum=1.0, maximum=15.0, step=0.5, label='CFG Scale', value=7.5, elem_id="i2v_cfg_scale")
+                    with gr.Row():
+                        i2v_steps = gr.Slider(minimum=1, maximum=50, step=1, elem_id="i2v_steps", label="Sampling steps", value=30)
+                        i2v_motion = gr.Slider(minimum=5, maximum=20, step=1, elem_id="i2v_motion", label="FPS", value=8)
+                    with gr.Row():
+                        i2v_video_length = gr.Slider(minimum=2, maximum=8, step=1, elem_id="i2v_video_length", label="Video Length (seconds)", value=2)
+                    i2v_end_btn = gr.Button("Generate")
+                with gr.Row():
+                    i2v_output_video = gr.Video(label="Generated Video",elem_id="output_vid",autoplay=True,show_share_button=True)
+
+            gr.Examples(examples=i2v_examples,
+                        inputs=[i2v_input_image, i2v_input_text, i2v_steps, i2v_cfg_scale, i2v_eta, i2v_motion, i2v_seed, i2v_video_length],
+                        outputs=[i2v_output_video],
+                        fn = infer,
+                        cache_examples=True,
+            )
+        i2v_end_btn.click(inputs=[i2v_input_image, i2v_input_text, i2v_steps, i2v_cfg_scale, i2v_eta, i2v_motion, i2v_seed, i2v_video_length],
+                        outputs=[i2v_output_video],
+                        fn = infer
+        )
+
+dynamicrafter_iface.queue(max_size=12).launch(show_api=True)