app.py

from transformers import AutoModel, AutoTokenizer
from flask import Flask, request, jsonify
import tensorflow as tf

app = Flask(__name__)

# Load Hugging Face model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("Erfan11/Neuracraft", use_auth_token="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")
hf_model = AutoModel.from_pretrained("Erfan11/Neuracraft", use_auth_token="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")

# Load TensorFlow model
tf_model = tf.keras.models.load_model('path_to_your_tf_model.h5')

@app.route('/predict', methods=['POST'])
def predict():
    data = request.get_json()
    
    # Tokenize the input using Hugging Face's tokenizer
    inputs = tokenizer(data["text"], return_tensors="pt")
    
    # Make prediction with Hugging Face model
    hf_outputs = hf_model(**inputs)
    
    # Optionally: You can also add TensorFlow model predictions here, depending on what it’s used for.
    # Assuming the TensorFlow model is used for something else like feature extraction
    tf_outputs = tf_model.predict([data["text"]])  # Modify based on your input processing

    return jsonify({
        "hf_outputs": hf_outputs[0].tolist(),  # Convert Hugging Face output to JSON serializable format
        "tf_outputs": tf_outputs.tolist()  # Convert TensorFlow output to JSON serializable format
    })

if __name__ == '__main__':
    app.run(debug=True)