Update app.py

app.py

@@ -32,11 +32,14 @@ print(f"Diretório atual adicionado ao sys.path.")
 # --- ETAPA 3: Instalar Dependências Corretamente ---
 python_executable = sys.executable
 
-# CORREÇÃO CRÍTICA: Filtrar requirements.txt para evitar conflitos com torch/torchvision
-print("Filtrando requirements.txt para evitar conflitos de versão...")
+# CORREÇÃO: Forçar uma versão do NumPy < 2.0 para evitar conflitos de compatibilidade.
+print("Instalando NumPy compatível...")
+subprocess.run([python_executable, "-m", "pip", "install", "numpy<2.0"], check=True)
+
+# Filtrar requirements.txt para evitar conflitos com torch/torchvision pré-instalados
+print("Filtrando requirements.txt...")
 with open("requirements.txt", "r") as f_in, open("filtered_requirements.txt", "w") as f_out:
     for line in f_in:
-        # Ignora as linhas que podem causar conflitos
         if not line.strip().startswith(('torch', 'torchvision')):
             f_out.write(line)
 
@@ -104,26 +107,18 @@ from data.image.transforms.na_resize import NaResize
 from data.video.transforms.rearrange import Rearrange
 from common.config import load_config
 from common.distributed import init_torch
-from common.distributed.advanced import init_sequence_parallel
 from common.seed import set_seed
-from common.partition import partition_by_size
 from projects.video_diffusion_sr.infer import VideoDiffusionInfer
 from common.distributed.ops import sync_data
 
-torch.hub.download_url_to_file('https://huggingface.co/datasets/Iceclear/SeedVR_VideoDemos/resolve/main/seedvr_videos_crf23/aigc1k/23_1_lq.mp4', '01.mp4')
-torch.hub.download_url_to_file('https://huggingface.co/datasets/Iceclear/SeedVR_VideoDemos/resolve/main/seedvr_videos_crf23/aigc1k/28_1_lq.mp4', '02.mp4')
-torch.hub.download_url_to_file('https://huggingface.co/datasets/Iceclear/SeedVR_VideoDemos/resolve/main/seedvr_videos_crf23/aigc1k/2_1_lq.mp4', '03.mp4')
-print("✅ Setup completo. Iniciando a aplicação...")
-
-
-
-
 os.environ["MASTER_ADDR"] = "127.0.0.1"
 os.environ["MASTER_PORT"] = "12355"
 os.environ["RANK"] = str(0)
 os.environ["WORLD_SIZE"] = str(1)
 
 use_colorfix = os.path.exists("projects/video_diffusion_sr/color_fix.py")
+if use_colorfix:
+    from projects.video_diffusion_sr.color_fix import wavelet_reconstruction
 
 def configure_runner():
     config = load_config('configs_3b/main.yaml')
@@ -155,11 +150,25 @@ def generation_step(runner, text_embeds_dict, cond_latents):
 def generation_loop(video_path, seed=666, fps_out=24):
     if video_path is None: return None, None, None
     runner = configure_runner()
-    text_embeds = {"texts_pos": [torch.load('pos_emb.pt').to("cuda")], "texts_neg": [torch.load('neg_emb.pt').to("cuda")]}
+    # Adicionado `weights_only=True` para segurança e para suprimir o aviso
+    text_embeds = {
+        "texts_pos": [torch.load('pos_emb.pt', weights_only=True).to("cuda")], 
+        "texts_neg": [torch.load('neg_emb.pt', weights_only=True).to("cuda")]
+    }
     runner.configure_diffusion()
     set_seed(int(seed))
     os.makedirs("output", exist_ok=True)
-    transform = Compose([NaResize(1024), DivisibleCrop(16), Normalize(0.5, 0.5), Rearrange("t c h w -> c t h w")])
+    
+    # CORREÇÃO: Fornecer os argumentos que faltam para NaResize.
+    res_h, res_w = 1280, 720
+    transform = Compose([
+        NaResize(resolution=(res_h * res_w)**0.5, mode="area", downsample_only=False),
+        Lambda(lambda x: torch.clamp(x, 0.0, 1.0)),
+        DivisibleCrop((16, 16)), 
+        Normalize(0.5, 0.5), 
+        Rearrange("t c h w -> c t h w")
+    ])
+
     media_type, _ = mimetypes.guess_type(video_path)
     is_video = media_type and media_type.startswith("video")
     
@@ -187,9 +196,7 @@ def generation_loop(video_path, seed=666, fps_out=24):
 
 with gr.Blocks(title="SeedVR") as demo:
     gr.HTML(f"""
-        <div style='text-align:center; margin-bottom: 10px;'>
-            <img src='file/{os.path.abspath("assets/seedvr_logo.png")}' style='height:40px;' alt='SeedVR logo'/>
-        </div>
+        
         <p><b>Demonstração oficial do Gradio</b> para 
         <a href='https://github.com/ByteDance-Seed/SeedVR' target='_blank'>
         <b>SeedVR2: One-Step Video Restoration via Diffusion Adversarial Post-Training</b></a>.<br>
@@ -206,11 +213,6 @@ with gr.Blocks(title="SeedVR") as demo:
     output_video = gr.Video(label="Vídeo de Saída")
     download_link = gr.File(label="Baixar Resultado")
     run_button.click(fn=generation_loop, inputs=[input_file, seed, fps], outputs=[output_image, output_video, download_link])
-    gr.Examples(examples=[["01.mp4", 42, 24], ["02.mp4", 42, 24], ["03.mp4", 42, 24]], inputs=[input_file, seed, fps])
-    gr.HTML("""
-        <hr>
-        <p>Se você achou o SeedVR útil, por favor ⭐ o 
-        <a href='https://github.com/ByteDance-Seed/SeedVR' target='_blank'>repositório no GitHub</a>.</p>
-    """)
+    
 
 demo.queue().launch(share=True)