app.py

# -*- coding: utf-8 -*-
"""app.ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/1FHgZJGBHIJ2C2OvYebeuCqVw1F7Rc29k
"""

from fastapi import FastAPI
from pydantic import BaseModel
import re
from underthesea import word_tokenize

import tensorflow as tf
from tensorflow.keras.layers import InputLayer
from tensorflow.keras.preprocessing.sequence import pad_sequences
import pickle

app = FastAPI()

# ---- PATCH InputLayer ----
_original_init = InputLayer.__init__

def _patched_init(self, *args, **kwargs):
    if "batch_shape" in kwargs and "batch_input_shape" not in kwargs:
        kwargs["batch_input_shape"] = kwargs.pop("batch_shape")
    return _original_init(self, *args, **kwargs)

InputLayer.__init__ = _patched_init
# ---- END PATCH ----

"""1. Load mô hình + tokenizer"""

# Load model
import tensorflow as tf
model = tf.keras.models.load_model("bilstm_model.h5")

# Load tokenizer
import pickle
with open("tokenizer.pkl", "rb") as f:
    tokenizer = pickle.load(f)

"""2. Làm sạch dữ liệu"""

import re

def clean_text(text):
    text = text.lower()
    text = re.sub(r"http\S+", "", text)
    text = re.sub(r"[^a-zA-Z0-9áàảãạăắằẳẵặâấầẩẫậéèẻẽẹêếềểễệíìỉĩịóòỏõọôốồổỗộơớờởỡợúùủũụưứừửữựýỳỷỹỵđ ]", " ", text)
    text = re.sub(r"\s+", " ", text).strip()
    text = word_tokenize(text, format="text")
    return text

class PredictRequest(BaseModel):
    text: str


class PredictResponse(BaseModel):
    label: int
    prob: float

"""2.5. Hàm test api"""

@app.post("/health")
def health():
    return {"status": "ok"}

"""3. Hàm dự đoán"""

max_len = 300

@app.post("/predict", response_model=PredictResponse)
def predict(req: PredictRequest):
    # 1. Clean
    cleaned = clean_text(req.text)

    # 2. Tokenize
    seq = tokenizer.texts_to_sequences([cleaned])

    # 3. Pad
    padded = pad_sequences(seq, maxlen=max_len)

    # 4. Predict
    prob = model.predict(padded)[0][0] #khi dự báo, kết quả trả về là 1 ma trận dù chỉ có 1 phần tử :))? in ra giá trị đầu của ma trận là có thể dùng như số(if else)
    label = 1 if prob > 0.5 else 0

    return PredictResponse(label=label, prob=prob)