Update app.py

app.py

@@ -5,11 +5,10 @@ from safetensors.torch import load_file
 from transformers import AutoTokenizer, AutoModel
 import gc
 
-# 清理内存
+# Release memory
 gc.collect()
 torch.cuda.empty_cache()
 
-# 1. 定义MultiTaskRoberta模型架构
 class MultiTaskRoberta(nn.Module):
     def __init__(self, base_model):
         super().__init__()
@@ -24,30 +23,27 @@ class MultiTaskRoberta(nn.Module):
         regs = self.regressor(pooled)
         return {"logits": logits, "regression_outputs": regs}
 
-# 2. 准备模型和tokenizer
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-print(f"使用设备: {device}")
+print(f"Device: {device}")
 
-# 加载tokenizer
+# Load tokenizer
 tokenizer = AutoTokenizer.from_pretrained("hfl/chinese-roberta-wwm-ext")
 
-# 加载模型
+# Load base model
 base_model = AutoModel.from_pretrained("hfl/chinese-roberta-wwm-ext")
 model = MultiTaskRoberta(base_model)
 
-# 加载权重
-model_path = "model.safetensors"
+# Load safetensors
+model_path = "model1.safetensors"
 state_dict = load_file(model_path, device="cpu")
 model.load_state_dict(state_dict)
 model.to(device)
 model.eval()
 
-# 使用半精度减少内存占用
+# Use half precision to reduce memory usage
 # if device.type == 'cuda':
 #     model.half()
-#     print("使用半精度模型")
 
-# 3. 优化后的推理函数
 def predict(text: str):
     try:
         inputs = tokenizer(
@@ -58,7 +54,6 @@ def predict(text: str):
             max_length=128
         )
 
-        # 将输入移到设备
         inputs = {k: v.to(device) for k, v in inputs.items()}
 
         with torch.no_grad():
@@ -71,7 +66,6 @@ def predict(text: str):
         pred_class = torch.argmax(out["logits"], dim=-1).item()
         sentiment_map = {0: "正面", 1: "負面", 2: "中立"}
 
-        # 将结果移回CPU处理
         reg_results = out["regression_outputs"][0].cpu().numpy()
         rating, delight, anger, sorrow, happiness = reg_results
 
@@ -86,7 +80,7 @@ def predict(text: str):
     except Exception as e:
         return {"错误": f"处理失败: {str(e)}"}
 
-# 4. 创建Gradio界面
+# Create Gradio interface
 iface = gr.Interface(
     fn=predict,
     inputs=gr.Textbox(lines=3, placeholder="請輸入粵語文本...", label="粵語文本"),
@@ -108,6 +102,5 @@ iface = gr.Interface(
     ]
 )
 
-# 5. 启动应用 - 使用兼容的启动方式
 if __name__ == "__main__":
     iface.launch(share=True, show_error=True)