chnage reqs

app.py

@@ -1,7 +1,6 @@
 from flask import Flask, request, jsonify
 from flask_cors import CORS
-import torch
-from transformers import GPT2Tokenizer, GPT2LMHeadModel
+from transformers import pipeline
 import os
 import re
 import logging
@@ -16,57 +15,40 @@ CORS(app)
 
 # Model loading
 MODEL_PATH = os.environ.get('MODEL_PATH', 'samdak93/qrit-2')
-tokenizer = None
-model = None
-device = None
+generator = None
 
 def load_model():
-    global tokenizer, model, device
+    global generator
     logger.info(f"Loading model from {MODEL_PATH}")
     try:
-        tokenizer = GPT2Tokenizer.from_pretrained(MODEL_PATH)
-        model = GPT2LMHeadModel.from_pretrained(MODEL_PATH)
+        # Use Hugging Face pipeline for text generation
+        generator = pipeline("text-generation", model=MODEL_PATH)
+        logger.info("Model loaded successfully")
     except Exception as e:
         logger.error(f"Error loading custom model: {e}")
         logger.info("Falling back to default GPT-2")
-        tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
-        model = GPT2LMHeadModel.from_pretrained("gpt2")
-
-    tokenizer.pad_token = tokenizer.eos_token
-    model.config.pad_token_id = model.config.eos_token_id
-
-    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-    model.to(device)
-    logger.info(f"Model loaded on device: {device}")
+        # Fallback to default GPT-2 if model loading fails
+        generator = pipeline("text-generation", model="gpt2")
+        logger.info("Default GPT-2 model loaded")
 
 def generate_recipe(goal_prompt, max_length=300, temperature=0.7):
-    global tokenizer, model, device
-    if tokenizer is None or model is None:
+    global generator
+    if generator is None:
         load_model()
 
     full_prompt = f"### Goal: {goal_prompt}\n### Recipe: "
-    input_ids = tokenizer.encode(full_prompt, return_tensors="pt").to(device)
-    attention_mask = torch.ones_like(input_ids).to(device)
-
-    with torch.no_grad():
-        output = model.generate(
-            input_ids=input_ids,
-            attention_mask=attention_mask,
-            max_length=max_length,
-            temperature=temperature,
-            do_sample=True,
-            top_k=50,
-            top_p=0.95,
-            no_repeat_ngram_size=2,
-            pad_token_id=tokenizer.eos_token_id
-        )
-
-    generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
-    recipe_part = generated_text[len(full_prompt):]
-    if "### Goal:" in recipe_part:
-        recipe_part = recipe_part.split("### Goal:")[0].strip()
-
-    return full_prompt + recipe_part
+    
+    # Generate recipe using the Hugging Face pipeline
+    try:
+        generated_text = generator(full_prompt, max_length=max_length, temperature=temperature, num_return_sequences=1)[0]['generated_text']
+        recipe_part = generated_text[len(full_prompt):]
+        if "### Goal:" in recipe_part:
+            recipe_part = recipe_part.split("### Goal:")[0].strip()
+
+        return full_prompt + recipe_part
+    except Exception as e:
+        logger.error(f"Error generating recipe: {str(e)}")
+        return str(e)
 
 def parse_recipe(recipe_text):
     recipe_data = {

requirements.txt

@@ -1,6 +1,6 @@
-flask==2.2.2  # Adjusted Flask version to a stable, widely used one
+flask==3.1.0
+flask_cors==5.0.1
 flask_cors==5.2.0  # Updated flask_cors to the latest version
 transformers==4.30.0  # Updated transformers version (compatible with pipeline)
 torch==2.1.0  # Updated torch to a stable version (optional based on need for GPU support)
 numpy<2  # No change to numpy
-pyngrok==5.1.0  # Optional if you use pyngrok for ngrok tunneling