Switch to CPU execution

app.py

@@ -9,7 +9,7 @@ model_path = "hiko1999/Qwen2-Wildfire-VL-2B-Instruct"  # 替换为你的模型
 
 # 加载 Hugging Face 上的模型和 processor
 tokenizer = AutoTokenizer.from_pretrained(model_path)
-model = Qwen2VLForConditionalGeneration.from_pretrained(model_path, torch_dtype=torch.bfloat16, device_map="auto")
+model = Qwen2VLForConditionalGeneration.from_pretrained(model_path, torch_dtype=torch.bfloat16)  # 移除 device_map 参数以避免自动分配到 GPU
 processor = AutoProcessor.from_pretrained(model_path)
 
 # 定义预测函数
@@ -22,7 +22,9 @@ def predict(image):
     text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
     image_inputs, video_inputs = process_vision_info(messages)
     inputs = processor(text=[text], images=image_inputs, videos=video_inputs, padding=True, return_tensors="pt")
-    inputs = inputs.to("cuda")  # 转移到GPU
+
+    # 将数据转移到 CPU
+    inputs = inputs.to("cpu")  # 使用 CPU 而不是 CUDA
 
     # 生成模型输出
     generated_ids = model.generate(**inputs, max_new_tokens=128)