Init commit

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+receipt_ner_model.keras filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,114 @@
+import os
+import re
+import pickle
+import numpy as np
+import tensorflow as tf
+from fastapi import FastAPI
+from pydantic import BaseModel
+from fastapi.middleware.cors import CORSMiddleware
+
+# Optimasi Memory (Wajib untuk HF Free Tier)
+os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
+os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
+os.environ["TF_NUM_INTEROP_THREADS"] = "1"
+os.environ["TF_NUM_INTRAOP_THREADS"] = "1"
+
+app = FastAPI(title="NER API Service")
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+# Global variables untuk Lazy Loading
+ner_model = None
+vocab = None
+
+# Konfigurasi NER
+MAX_SEQ_LEN = 64
+IDX2LABEL = {0: "<PAD>", 1: "O", 2: "B-MERCHANT", 3: "I-MERCHANT", 4: "B-ITEM", 5: "I-ITEM", 6: "B-TOTAL", 7: "I-TOTAL"}
+UNK_TOKEN = "<UNK>"
+NUM_TOKEN = "<NUM>"
+PRICE_TOKEN = "<PRICE>"
+
+# Schema untuk menerima data JSON dari Space A
+class TextRequest(BaseModel):
+    text: str
+
+def get_ner_model():
+    global ner_model, vocab
+    if ner_model is None:
+        with open("data/processed/vocab.pkl", "rb") as f:
+            vocab = pickle.load(f)
+        # compile=False penting agar tidak error saat mencari custom loss function
+        ner_model = tf.keras.models.load_model("receipt_ner_model.keras", compile=False)
+    return ner_model, vocab
+
+def normalize_token(token: str):
+    token = token.strip()
+    if re.fullmatch(r"\d{1,3}([.,]\d{3})*", token):
+        return PRICE_TOKEN
+    if re.fullmatch(r"\d+", token):
+        return NUM_TOKEN
+    token = re.sub(r"^[^a-zA-Z0-9']+|[^a-zA-Z0-9']+$", "", token)
+    return token.upper() if token else UNK_TOKEN
+
+@app.post("/predict-ner/")
+def predict_ner(request: TextRequest):
+    try:
+        raw_text = request.text
+        entities = {"MERCHANT": [], "ITEM": [], "TOTAL": []}
+        
+        # Jika teks kosong, langsung kembalikan entitas kosong
+        if not raw_text.strip():
+            return {"success": True, "entities": entities}
+
+        model, vcb = get_ner_model()
+        
+        # Tokenisasi sederhana
+        tokens = [t for line in raw_text.split("\n") for t in line.strip().split() if t]
+        if not tokens:
+            return {"success": True, "entities": entities}
+
+        # Encoding
+        norm_tokens = [normalize_token(t) for t in tokens]
+        X = np.zeros((1, MAX_SEQ_LEN), dtype=np.int32)
+        mask = np.zeros((1, MAX_SEQ_LEN), dtype=np.float32)
+        
+        length = min(len(norm_tokens), MAX_SEQ_LEN)
+        for i in range(length):
+            X[0, i] = vcb.get(norm_tokens[i], vcb.get(UNK_TOKEN, 0))
+            mask[0, i] = 1.0
+            
+        # Prediksi menggunakan TensorFlow
+        probs = model([X, mask], training=False).numpy()
+        labels = [IDX2LABEL[int(np.argmax(probs[0, i]))] for i in range(length)]
+        
+        # Ekstraksi IOB (Inside-Outside-Beginning)
+        current_type = None
+        span = []
+        for tok, lbl in zip(tokens[:length], labels):
+            if lbl.startswith("B-"):
+                if current_type: 
+                    entities[current_type].append(" ".join(span))
+                current_type = lbl[2:]
+                span = [tok]
+            elif lbl.startswith("I-") and current_type == lbl[2:]:
+                span.append(tok)
+            else:
+                if current_type: 
+                    entities[current_type].append(" ".join(span))
+                current_type = None
+                span = []
+        if current_type: 
+            entities[current_type].append(" ".join(span))
+
+        return {"success": True, "entities": entities}
+        
+    except Exception as e:
+        print(f"Error NER: {e}")
+        # Jika terjadi error internal, kembalikan objek kosong agar Space A tidak crash
+        return {"success": False, "error": str(e), "entities": {"MERCHANT": [], "ITEM": [], "TOTAL": []}}

receipt_ner_model.keras

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b91e1ef3320864814b5dfda22768766247b98f4246bc64a63c870ab0ed86e737
+size 10616552

requirements.txt

@@ -0,0 +1,5 @@
+fastapi
+uvicorn
+tensorflow>=2.16.1
+pydantic
+numpy