Update app.py

app.py

@@ -6,7 +6,6 @@ import os
 import joblib
 from model import DualStreamTransformer, ArcMarginProduct
 
-
 DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 FOLD = 5
 MODEL_PATH = f"best_model_fold_{FOLD}.pt"
@@ -17,26 +16,105 @@ metric_fc = ArcMarginProduct(32, 2).to(DEVICE)
 
 if os.path.exists(MODEL_PATH):
     checkpoint = torch.load(MODEL_PATH, map_location=DEVICE)
-    model.load_state_dict(checkpoint['model'])
-    metric_fc.load_state_dict(checkpoint['fc'])
+    model.load_state_dict(checkpoint['model'], strict=False)
+    metric_fc.load_state_dict(checkpoint['fc'], strict=False)
     model.eval()
-    print(" Model loaded successfully")
+
 
 scaler_ccmq = joblib.load(f"scaler_ccmq_fold_{FOLD}.pkl")
 scaler_osdi = joblib.load(f"scaler_osdi_fold_{FOLD}.pkl")
 
 def analyze_and_predict(*all_answers):
-
-    print(f"DEBUG: Received answers count = {len(all_answers)}")
-    
     ccmq_map = {"總是": 5, "經常": 4, "有時": 3, "很少": 2, "沒有": 1}
     osdi_map = {"總是": 4, "經常": 3, "一半一半": 2, "偶而": 1, "完全不曾": 0}
 
     try:
-
         x1_vals = [ccmq_map.get(a, 1) for a in all_answers[:24]]
         x2_vals = [osdi_map.get(a, 0) for a in all_answers[24:34]]
         
+        x1_scaled = scaler_ccmq.transform(np.array([x1_vals]))
+        x2_scaled = scaler_osdi.transform(np.array([x2_vals]))
+        
+        sx1 = torch.tensor(x1_scaled, dtype=torch.float32).to(DEVICE)
+        sx2 = torch.tensor(x2_scaled, dtype=torch.float32).to(DEVICE)
+
+        with torch.no_grad():
+            feats = model(sx1, sx2)
+            logits = metric_fc.predict(feats)
+            probs = torch.softmax(logits, dim=1)
+            pred_idx = torch.argmax(probs, dim=1).item()
+            conf = probs[0, pred_idx].item()
+    except Exception as e:
+        raise gr.Error(f"推論出錯: {e}")
+
+    res_label = " 乾眼風險 (SJS/DES)" if pred_idx == 1 else " 正常/健康"
+    table_data = [[f"項目 {i+1}", str(all_answers[i]), "OK"] for i in range(len(all_answers))]
+    
+ 
+    return (
+        gr.update(visible=False), 
+        gr.update(visible=True),  
+        f"### {res_label}", 
+        f"AI 推論成功！診斷信心度為 {conf:.2%}。分析已整合 CCMQ 體質與 OSDI 臨床特徵。",
+        {"Risk": conf if pred_idx==1 else 1-conf, "Healthy": 1 - (conf if pred_idx==1 else 1-conf)},
+        table_data
+    )
+
+with gr.Blocks(css=".scroll-box { height: 400px; overflow-y: auto; border: 1px solid #ddd; padding: 10px; }") as demo:
+    gr.Markdown("#  中醫 AI 多模態診斷系統")
+
+ 
+    with gr.Column(visible=True) as input_area:
+        with gr.Row():
+            with gr.Column():
+                gr.Markdown("#### Step 1: CCMQ 體質量表")
+                with gr.Group(elem_classes="scroll-box"):
+                    ccmq_labels = ["惡寒惡風", "自汗", "胸悶腹脹","咽喉痰梗感","多愁善感","易受驚","面部暗沉","黑眼圈","健忘","唇色暗","身熱、面熱","膚乾口乾","唇紅","便祕","兩顴紅","眼乾澀","四肢冷","惡寒、腰膝冷","飲冷腹瀉","口苦口臭","帶下色黃/下陰潮濕","鼻塞流涕","變天咳喘","過敏"]
+                    all_ccmq = [gr.Radio(["總是", "經常", "有時", "很少", "沒有"], label=f"{i+1}. {txt}", value="沒有") for i, txt in enumerate(ccmq_labels)]
+            
+            with gr.Column():
+                gr.Markdown("#### Step 2: OSDI 症狀量表")
+                with gr.Group(elem_classes="scroll-box"):
+                    osdi_labels = ["1. 對光敏感", "2. 眼睛疼痛", "3. 視線模糊", "4. 視力減退", "5. 閱讀限制", "6. 夜間駕駛", "7. 電腦操作", "8. 觀看電視", "9. 刮風不適", "10. 空調不適"]
+                    all_osdi = [gr.Radio(["總是", "經常", "一半一半", "偶而", "完全不曾"], label=txt, value="完全不曾") for txt in osdi_labels]
+                
+                submit_btn = gr.Button("提交診斷並生成報告", variant="primary", size="lg")
+
+
+    with gr.Column(visible=False) as result_area:
+        gr.Markdown("### AI 診斷分析報告")
+        res_title = gr.Markdown("### 結論加載中...")
+        with gr.Row():
+            with gr.Column():
+                res_prob = gr.Label(label="預測機率分佈")
+                res_desc = gr.Markdown("報告詳情...")
+            with gr.Column():
+                res_table = gr.Dataframe(headers=["項目", "回答", "狀態"], interactive=False)
+        
+        reset_btn = gr.Button("重新開始測驗", variant="secondary")
+
+    all_inputs = all_ccmq + all_osdi
+    
+    submit_btn.click(
+        fn=analyze_and_predict, 
+        inputs=all_inputs, 
+        outputs=[input_area, result_area, res_title, res_desc, res_prob, res_table]
+    )
+    
+    reset_btn.click(
+        fn=lambda: [gr.update(visible=True), gr.update(visible=False)] + ["沒有"]*24 + ["完全不曾"]*10,
+        outputs=[input_area, result_area] + all_inputs
+    )
+
+if __name__ == "__main__":
+
+    demo.launch(
+        ssr_mode=False,
+        show_api=False,
+        quiet=True
+    )        x1_vals = [ccmq_map.get(a, 1) for a in all_answers[:24]]
+        x2_vals = [osdi_map.get(a, 0) for a in all_answers[24:34]]
+        
         x1_raw = np.array([x1_vals])
         x2_raw = np.array([x2_vals])
 
@@ -75,7 +153,7 @@ def analyze_and_predict(*all_answers):
         None
     )
 
-# --- UI 介面 ---
+
 with gr.Blocks() as demo:
     gr.Markdown("# 中醫 AI 診斷系統")