Update app.py

app.py

@@ -251,6 +251,7 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 # URL do config.json no Space
 config_url = "https://huggingface.co/spaces/vcollos/family/resolve/main/config.json"
 local_config_path = "./config.json"
+weights_path = "./ae.safetensors"
 
 # Baixar o config.json, se ainda não existir localmente
 if not os.path.exists(local_config_path):
@@ -259,13 +260,9 @@ if not os.path.exists(local_config_path):
     with open(local_config_path, "wb") as f:
         f.write(response.content)
 
-# Caminho para os pesos
-base_model = "SG161222/Verus_Vision_1.0b"
-weights_path = "./ae.safetensors"
-
 # Baixar os pesos do modelo se necessário
 if not os.path.exists(weights_path):
-    weights_url = f"https://huggingface.co/{base_model}/resolve/main/ae.safetensors"
+    weights_url = f"https://huggingface.co/SG161222/Verus_Vision_1.0b/resolve/main/ae.safetensors"
     response = requests.get(weights_url)
     response.raise_for_status()
     with open(weights_path, "wb") as f:
@@ -273,21 +270,22 @@ if not os.path.exists(weights_path):
 
 # Carregar o Autoencoder com o config.json local
 good_vae = AutoencoderKL.from_pretrained(
-    pretrained_model_name_or_path=None,  # Não buscar no repositório
-    config=local_config_path,            # Configuração local
-    filename=weights_path,               # Pesos locais
+    pretrained_model_name_or_path=local_config_path,  # Caminho local do config.json
+    filename=weights_path,  # Pesos locais
     torch_dtype=dtype
 ).to(device)
 
-# Carregar o pipeline principal com o config.json local
+# Configurar o pipeline principal com o config.json local
 pipe = DiffusionPipeline.from_pretrained(
-    pretrained_model_name_or_path=None,  # Não buscar no repositório
-    config=local_config_path,            # Configuração local
+    pretrained_model_name_or_path=local_config_path,  # Caminho local do config.json
     filename="VerusVision_1.0b_Transformer_fp16.safetensors",
     torch_dtype=dtype
 ).to(device)
 
 print("Modelo carregado com sucesso!")
+
+
+
 # Se precisar do Image-to-Image
 pipe_i2i = AutoPipelineForImage2Image.from_pretrained(
     base_model,