Update app.py

app.py

@@ -1,6 +1,8 @@
 import os
 import re
 import pickle
+import json
+import zipfile
 import numpy as np
 import tensorflow as tf
 from fastapi import FastAPI
@@ -38,15 +40,54 @@ class TextRequest(BaseModel):
     text: str
 
 # =========================================================
-# LAZY LOAD SANGAT BERSIH (Tanpa Bypass)
+# OPERASI BEDAH FILE .KERAS (Solusi Definitif Error JSON)
+# =========================================================
+def patch_keras_model(original_path):
+    patched_path = "patched_" + original_path
+    # Jika sudah pernah dibedah, langsung gunakan hasilnya
+    if os.path.exists(patched_path):
+        return patched_path
+    
+    print(f"Membedah {original_path} untuk menghapus quantization_config...")
+    try:
+        with zipfile.ZipFile(original_path, 'r') as zin:
+            with zipfile.ZipFile(patched_path, 'w') as zout:
+                for item in zin.infolist():
+                    content = zin.read(item.filename)
+                    if item.filename == 'config.json':
+                        # Bongkar JSON-nya
+                        config_dict = json.loads(content.decode('utf-8'))
+                        
+                        # Fungsi rekursif untuk mencari dan menghapus quantization_config di semua layer
+                        def hapus_biang_kerok(d):
+                            if isinstance(d, dict):
+                                d.pop('quantization_config', None)
+                                for k, v in d.items():
+                                    hapus_biang_kerok(v)
+                            elif isinstance(d, list):
+                                for i in d:
+                                    hapus_biang_kerok(i)
+                        
+                        hapus_biang_kerok(config_dict)
+                        # Bungkus kembali jadi JSON
+                        content = json.dumps(config_dict).encode('utf-8')
+                        
+                    zout.writestr(item, content)
+        return patched_path
+    except Exception as e:
+        print(f"Gagal membedah model: {e}")
+        return original_path # Fallback ke file asli jika gagal
+
+# =========================================================
+# LAZY LOAD MENGGUNAKAN MODEL YANG SUDAH DIBEDAH
 # =========================================================
 def get_ner_model():
     global ner_model, vocab
     if ner_model is None:
         with open("vocab.pkl", "rb") as f:
             vocab = pickle.load(f)
-        # Langsung load utuh! Keras 3 sudah aktif.
-        ner_model = tf.keras.models.load_model("receipt_ner_model.keras", compile=False)
+        patched_file = patch_keras_model("receipt_ner_model.keras")
+        ner_model = tf.keras.models.load_model(patched_file, compile=False)
     return ner_model, vocab
 
 def get_clf_model():
@@ -54,8 +95,8 @@ def get_clf_model():
     if clf_model is None:
         with open('penerjemah_kategori.pkl', 'rb') as file:
             clf_encoder = pickle.load(file)
-        # Langsung load utuh! Kamus TextVectorization akan otomatis ikut termuat.
-        clf_model = tf.keras.models.load_model('model_strukly.keras', compile=False)
+        patched_file = patch_keras_model("model_strukly.keras")
+        clf_model = tf.keras.models.load_model(patched_file, compile=False)
     return clf_model, clf_encoder
 
 def normalize_token(token: str):
@@ -89,7 +130,6 @@ def predict_ner(request: TextRequest):
                 X[0, i] = vcb.get(norm_tokens[i], vcb.get(UNK_TOKEN, 0))
                 mask[0, i] = 1.0
                 
-            # Kembali menggunakan mask karena model di-load utuh
             probs = model([X, mask], training=False).numpy()
             labels = [IDX2LABEL[int(np.argmax(probs[0, i]))] for i in range(length)]