Update app.py

app.py

@@ -2,6 +2,7 @@ import torch
 from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
 from datasets import load_dataset
 import spacy
+import gradio as gr
 
 # 設置設備和環境變數（如有需要）
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
@@ -21,57 +22,57 @@ whisper_pipe = pipeline(
     feature_extractor=whisper_processor.feature_extractor,
     device=device)
 
-# 安裝必要庫並下載 spaCy 英文小模型（如果尚未安裝）
+# DeepSeek-V3 模型初始化（文本生成）
+deepseek_pipe = None  # 預設值，以防模型加載失敗
+
 try:
     
+        # 這裡需要添加正確的內容以避免 IndentationError
+        
+        deepseek_pipe=pipeline("text-generation",model="t5-base")  # 暫時使用 T5 作為替代
+        
 except Exception as e:
     
-finally:
-    
-spacy.cli.download("en_core_web_sm")
+        print(f"加載模型時出現錯誤：{e}")
 
-nlp = spacy.load("en_core_web_sm")
+# spaCy 初始化（文本分類與標籤）
+nlp=None 
 
-def process_audio(audio_file):
-    
+try:
         
+            nlp=spacy.load("en_core_web_sm")
             
-            # 語音轉文字
+except Exception as e:
             
-            result= whisper_pipe(audio_file)["text"]
+            print(f"加載spaCy模型時出現錯誤：{e}")
+
+def process_audio(audio_file):
+    
+        
             
+                result= whisper_pipe(audio_file)["text"]
                 
-                # 使用其他文本生成模型替換，因為目前無法直接加載DeepSeek-V3
-                
-                messages=[{"role": "user", "content": result}]
-                
-                deepseek_response="" # 預設回應
-                
-                try: 
-                    from transformers import pipeline
-                    
-                    pipe=pipeline("text-generation",model="t5-base")
-                    
-                    deepseek_response=pipe(messages)[0]["generated_text"]
                     
+                        messages=[{"role": "user", "content": result}]
+                        deepseek_response="" 
                         
+                        try: 
+                            if deepseek_pipe is not None:  
+                                deepseek_response=deepseek_pipe(messages)[0]["generated_text"]
+                            
+                            doc=nlp(deepseek_response) if nlp is not None else None 
+                            entities=[(ent.text, ent.label_) for ent in doc.ents] if doc is not None else []
+                            
+                            return {
+                                "Transcription (Whisper)": result,
+                                "AI Response (T5)": deepseek_response,# 修改為 T5 回應以避免與原來不同步                            
+                                "Extracted Entities (spaCy)": entities}
                         
-                        # 使用 spaCy 分析文本
-                        
-                        doc=nlp(deepseek_response)
-                        entities=[(ent.text, ent.label_) for ent in doc.ents]
-                        
-                        return {
-                            "Transcription (Whisper)": result,
-                            "AI Response (T5)": deepseek_response,# 修改為 T5 回應以避免與原來不同步                            
-                            "Extracted Entities (spaCy)": entities}
-                    
-                except Exception as e: 
-                    return {
-                        "Transcription (Whisper)": result,# 保留原始轉錄內容
-                        
-                        }
-                    
+                        except Exception as e: 
+                            return {
+                                "Transcription (Whisper)": result,# 保留原始轉錄內容                                
+                                    }
+                            
 
 
 with gr.Blocks() as app: