Update app.py

app.py

@@ -1,8 +1,6 @@
 import os
 import re
 import pickle
-import json
-import zipfile
 import numpy as np
 import tensorflow as tf
 from fastapi import FastAPI
@@ -40,7 +38,7 @@ class TextRequest(BaseModel):
     text: str
 
 # =========================================================
-# JURUS 1: BYPASS ARSITEKTUR MANUAL UNTUK NER (Anti-Error LSTM)
+# JURUS 1: BYPASS NER 
 # =========================================================
 def build_manual_ner_model(vocab_size=7929, max_len=64, num_classes=8):
     inputs = tf.keras.Input(shape=(max_len,), dtype='int32', name='tokens')
@@ -68,49 +66,49 @@ def build_manual_ner_model(vocab_size=7929, max_len=64, num_classes=8):
 def get_ner_model():
     global ner_model, vocab
     if ner_model is None:
-        with open("vocab.pkl", "rb") as f:
+        with open("data/processed/vocab.pkl", "rb") as f:
             vocab = pickle.load(f)
-        # Load kerangka manual, isi dengan weights
         ner_model = build_manual_ner_model()
         ner_model.load_weights("receipt_ner_model.keras")
     return ner_model, vocab
 
 # =========================================================
-# JURUS 2: BEDAH JSON UNTUK KLASIFIKASI (Menyelamatkan Kamus Kata)
+# JURUS 2: BYPASS KLASIFIKASI + PAKSA BANGUN KAMUS
 # =========================================================
-def patch_keras_model(original_path):
-    patched_path = "patched_" + original_path
-    if os.path.exists(patched_path): return patched_path
+def build_manual_clf_model():
+    inputs = tf.keras.Input(shape=(1,), dtype=tf.string, name='input_layer')
+    x = tf.keras.layers.TextVectorization(
+        max_tokens=5000, 
+        output_sequence_length=50, 
+        name='text_vectorization'
+    )(inputs)
+    x = tf.keras.layers.Embedding(input_dim=5000, output_dim=64, name='embedding')(x)
+    x = tf.keras.layers.GlobalAveragePooling1D(name='global_average_pooling1d')(x)
+    x = tf.keras.layers.Dense(64, activation='relu', name='dense')(x)
+    x = tf.keras.layers.Dropout(0.2, name='dropout')(x)
+    outputs = tf.keras.layers.Dense(7, activation='softmax', name='dense_1')(x)
     
-    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':
-                        config_dict = json.loads(content.decode('utf-8'))
-                        def hapus_biang_kerok(d):
-                            if isinstance(d, dict):
-                                d.pop('quantization_config', None)
-                                for v in d.values(): hapus_biang_kerok(v)
-                            elif isinstance(d, list):
-                                for i in d: hapus_biang_kerok(i)
-                        hapus_biang_kerok(config_dict)
-                        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
+    return tf.keras.Model(inputs=inputs, outputs=outputs, name='functional')
 
 def get_clf_model():
     global clf_model, clf_encoder
     if clf_model is None:
         with open('penerjemah_kategori.pkl', 'rb') as file:
             clf_encoder = pickle.load(file)
-        # Bedah JSON lalu load utuh agar TextVectorization bekerja
-        patched_file = patch_keras_model("model_strukly.keras")
-        clf_model = tf.keras.models.load_model(patched_file, compile=False)
+            
+        clf_model = build_manual_clf_model()
+        clf_model.load_weights("model_strukly.keras")
+        
+        # --- KODE AJAIB ANTI "Table not initialized" ---
+        tv_layer = clf_model.get_layer('text_vectorization')
+        weights = tv_layer.get_weights()
+        if weights:
+            # Ubah data bytes dari model kembali menjadi teks
+            vocab_list = [w.decode('utf-8') if isinstance(w, bytes) else str(w) for w in weights[0]]
+            # Paksa Keras membangun ulang tabelnya (Abaikan 2 token pertama: PAD dan UNK)
+            tv_layer.set_vocabulary(vocab_list[2:])
+        # -----------------------------------------------
+        
     return clf_model, clf_encoder
 
 # =========================================================
@@ -139,7 +137,7 @@ def predict_ner(request: TextRequest):
         
         if tokens:
             norm_tokens = [normalize_token(t) for t in tokens]
-            # Karena pakai arsitektur manual, mask tidak perlu dibuat eksplisit (otomatis di layer Embedding)
+            # Inference NER pakai manual bypass, tidak perlu mask eksplisit
             X = np.zeros((1, MAX_SEQ_LEN), dtype=np.int32)
             length = min(len(norm_tokens), MAX_SEQ_LEN)