Nafise
Flask pp
2f49beb
raw
history blame
2.38 kB
from flask import Flask, jsonify, request
import requests
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
app = Flask(__name__)
# Initialize sentiment analysis model
sentiment_tokenizer = AutoTokenizer.from_pretrained("mrm8488/t5-base-finetuned-emotion")
sentiment_model = AutoModelForSeq2SeqLM.from_pretrained("mrm8488/t5-base-finetuned-emotion")
# Initialize dialogue generation model
tokenizer = AutoTokenizer.from_pretrained("microsoft/GODEL-v1_1-large-seq2seq")
model = AutoModelForSeq2SeqLM.from_pretrained("microsoft/GODEL-v1_1-large-seq2seq")
# Last.fm API key
API_KEY = "e554f25da26e93055f2780bbe2b9293b"
# Function to generate response
def generate_response(dialog):
knowledge = ''
instruction = f'Instruction: given a dialog context, you need to respond empathically.'
dialog_text = ' EOS '.join(dialog)
query = f"{instruction} [CONTEXT] {dialog_text} {knowledge}"
input_ids = tokenizer.encode(query, return_tensors="pt")
output = model.generate(input_ids, max_length=16, min_length=2, top_p=0.9, do_sample=True)
generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
return generated_text
# Function to perform sentiment analysis
def sentiment_finder(user_dialog):
input_ids = sentiment_tokenizer.encode(user_dialog + '</s>', return_tensors='pt')
output = sentiment_model.generate(input_ids=input_ids, max_length=2)
emotion = [sentiment_tokenizer.decode(ids) for ids in output][0]
return emotion[6:]
@app.route("/get_response", methods=["POST", "GET"])
def get_response():
data = request.json
dialog = data.get('dialog', [])
generated_text = generate_response(dialog)
user_dialog = dialog[-1]
emotion = sentiment_finder(user_dialog)
# Fetch music recommendations based on emotion
recommendations_url = f"http://ws.audioscrobbler.com/2.0/?method=tag.gettoptracks&tag={emotion}&api_key={API_KEY}&format=json&limit=4"
recommendations_response = requests.get(recommendations_url)
recommendations = []
if recommendations_response.ok:
recommendations_data = recommendations_response.json()
recommendations = recommendations_data["tracks"]["track"]
response_data = {'generated_response': generated_text, 'recommendations': recommendations}
return jsonify(response_data)
if __name__ == '__main__':
app.run(port=8000)