Update app.py

app.py

@@ -2,44 +2,29 @@ from flask import Flask, request, jsonify
 import os
 import logging
 
-# Set up logging
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
 app = Flask(__name__)
-
-# Global variables for model
 model = None
-tokenizer = None
 
 def load_model():
-    global model, tokenizer
+    global model
     try:
-        logger.info("Loading MLX model...")
-        from mlx_lm import load, generate as mlx_generate
+        logger.info("Loading model with transformers...")
+        from transformers import pipeline
         
-        # Load your fine-tuned model
-        model, tokenizer = load(
-            'microsoft/Phi-3.5-mini-instruct',
-            adapter_path='kacperbb/phi-3.5-mlx-finetuned'
-        )
+        # Use the base model
+        model = pipeline("text-generation", model="microsoft/Phi-3.5-mini-instruct", trust_remote_code=True)
         logger.info("✅ Model loaded successfully!")
         return True
     except Exception as e:
-        logger.error(f"❌ Error loading MLX model: {e}")
-        try:
-            # Fallback to transformers
-            from transformers import pipeline
-            model = pipeline("text-generation", model="microsoft/Phi-3.5-mini-instruct")
-            logger.info("✅ Fallback model loaded!")
-            return True
-        except Exception as e2:
-            logger.error(f"❌ Fallback failed: {e2}")
-            return False
+        logger.error(f"❌ Error loading model: {e}")
+        return False
 
 @app.route('/generate', methods=['POST'])
 def generate_text():
-    global model, tokenizer
+    global model
     try:
         data = request.json
         prompt = data.get('inputs', data.get('prompt', ''))
@@ -48,13 +33,8 @@ def generate_text():
         if not prompt:
             return jsonify({"error": "No prompt provided"}), 400
         
-        if model and tokenizer:
-            # Use MLX
-            from mlx_lm import generate as mlx_generate
-            response = mlx_generate(model, tokenizer, prompt=prompt, max_tokens=max_tokens)
-        elif model:
-            # Use transformers pipeline
-            result = model(prompt, max_length=len(prompt.split()) + max_tokens, do_sample=True)
+        if model:
+            result = model(prompt, max_new_tokens=max_tokens, do_sample=True, temperature=0.7)
             response = result[0]['generated_text']
         else:
             return jsonify({"error": "Model not loaded"}), 500
@@ -68,7 +48,7 @@ def generate_text():
 @app.route('/health', methods=['GET'])
 def health():
     return jsonify({
-        "status": "healthy",
+        "status": "healthy", 
         "model_loaded": model is not None
     })
 
@@ -77,7 +57,7 @@ def home():
     return jsonify({
         "message": "Phi 3.5 API is running!",
         "endpoints": {
-            "generate": "POST /generate",
+            "generate": "POST /generate", 
             "health": "GET /health"
         }
     })