Update app.py

app.py

@@ -3,7 +3,7 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
 from accelerate import Accelerator
 
-# Check if GPU is available for better performancee
+# Check if GPU is available for better performance
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 print(f"Using device: {device}")
 
@@ -12,9 +12,9 @@ accelerator = Accelerator()
 
 # Load models and tokenizers with FP16 for speed optimization if GPU is available
 model_dirs = [
-    "Poonawala/gpt2",
-    "Poonawala/MiriFur",
-    "Poonawala/Llama-3.2-1B"
+    "muhammadAhmed22/fine_tuned_gpt2",
+    "muhammadAhmed22/MiriFurgpt2-recipes",
+    "muhammadAhmed22/auhide-chef-gpt-en"
 ]
 
 models = {}
@@ -23,10 +23,10 @@ tokenizers = {}
 def load_model(model_dir):
     model = AutoModelForCausalLM.from_pretrained(model_dir, torch_dtype=torch.float16 if device.type == "cuda" else torch.float32)
     tokenizer = AutoTokenizer.from_pretrained(model_dir)
-
+    
     if tokenizer.pad_token is None:
         tokenizer.pad_token = tokenizer.eos_token
-
+    
     # Move model to GPU/CPU as per availability
     model = model.to(device)
     return model, tokenizer
@@ -54,16 +54,16 @@ for model_dir in model_dirs:
 def get_response(prompt, model_name, user_type):
     if model_name not in models:
         return "Model not loaded correctly."
-
+    
     model = models[model_name]
     tokenizer = tokenizers[model_name]
 
     # Define different prompt templates based on user type
     user_type_templates = {
-        "Expert": f"As an Expert, {prompt}\nAnswer:",
-        "Intermediate": f"As an Intermediate, {prompt}\nAnswer:",
+        "Professional Chef": f"As a professional chef, {prompt}\nAnswer:",
+        "Home Cook": f"As a home cook, {prompt}\nAnswer:",
         "Beginner": f"Explain in simple terms: {prompt}\nAnswer:",
-        "Professional": f"As a Professional, {prompt}\nAnswer:"
+        "Food Enthusiast": f"As a food enthusiast, {prompt}\nAnswer:"
     }
 
     # Get the appropriate prompt based on user type
@@ -74,20 +74,21 @@ def get_response(prompt, model_name, user_type):
         return_tensors='pt',
         padding=True,
         truncation=True,
-        max_length=500  # Increased length for larger inputs
+        max_length=300  # Increased length for larger inputs
     ).to(device)
 
-    max_new_tokens = 200  # Increased to allow full-length answers
+    # Reduce max_new_tokens for faster response time
+    max_new_tokens = 50  # Reduced to speed up response time
 
     with torch.no_grad():
         output = model.generate(
             input_ids=encoding['input_ids'],
             attention_mask=encoding['attention_mask'],
-            max_new_tokens=max_new_tokens,  # Higher value for longer answers
-            num_beams=3,         # Using beam search for better quality answers
-            repetition_penalty=1.2,  # Increased to reduce repetitive text
-            temperature=0.9,     # Slightly higher for creative outputs
-            top_p=0.9,           # Including more tokens for diverse generation
+            max_new_tokens=max_new_tokens,
+            num_beams=1,         # Using greedy decoding (faster)
+            repetition_penalty=1.1,
+            temperature=0.7,     # Slightly reduced for better performance
+            top_p=0.85,          # Reduced top_p for faster results
             early_stopping=True,
             pad_token_id=tokenizer.pad_token_id
         )
@@ -99,65 +100,106 @@ def process_input(prompt, model_name, user_type):
     if prompt and prompt.strip():
         return get_response(prompt, model_name, user_type)
     else:
-        return "Please provide a prompt."
+        return "Please provide a valid prompt."
 
 # Gradio Interface with Modern Design
-with gr.Blocks(css="""
+with gr.Blocks(css=""" 
 body {
-    background-color: #faf3e0; /* Beige for a warm food-related theme */
-    font-family: 'Arial, sans-serif';
+    background-color: #f8f8f8;
+    font-family: 'Helvetica Neue', Arial, sans-serif;
 }
+
 .title {
-    font-size: 2.5rem;
-    font-weight: bold;
-    color: #ff7f50; /* Coral color for a food-inspired look */
+    font-size: 2.6rem;
+    font-weight: 700;
+    color: #ff6347;
     text-align: center;
-    margin-bottom: 1rem;
+    margin-bottom: 1.5rem;
 }
+
 .container {
-    max-width: 900px;
+    max-width: 800px;
     margin: auto;
     padding: 2rem;
     background-color: #ffffff;
     border-radius: 10px;
-    box-shadow: 0 4px 15px rgba(0, 0, 0, 0.1);
+    box-shadow: 0 12px 24px rgba(0, 0, 0, 0.1);
 }
+
 .button {
-    background-color: #ff7f50; /* Coral color for buttons */
+    background-color: #ff6347;
     color: white;
-    padding: 0.8rem 1.5rem;
-    font-size: 1rem;
+    padding: 0.8rem 1.8rem;
+    font-size: 1.1rem;
     border: none;
-    border-radius: 5px;
+    border-radius: 8px;
     cursor: pointer;
+    transition: background-color 0.3s ease;
+    margin-top: 1.5rem;
+    width: 100%;
 }
+
 .button:hover {
-    background-color: #ffa07a; /* Light salmon for hover effect */
+    background-color: #ff4500;
+}
+
+.gradio-interface .gr-textbox {
+    margin-bottom: 1.5rem;
+    width: 100%;
+    border-radius: 8px;
+    padding: 1rem;
+    border: 1px solid #ddd;
+    font-size: 1rem;
+    background-color: #f9f9f9;
+    color: #333;
+}
+
+.gradio-interface .gr-radio, .gradio-interface .gr-dropdown {
+    margin-bottom: 1.5rem;
+    width: 100%;
+    border-radius: 8px;
+    padding: 1rem;
+    border: 1px solid #ddd;
+    background-color: #f9f9f9;
+    font-size: 1rem;
+    color: #333;
+}
+
+.gradio-interface .gr-textbox[readonly] {
+    background-color: #f5f5f5;
+    color: #333;
+    font-size: 1rem;
 }
 """) as demo:
 
-    gr.Markdown("<div class='title'>Cookspert: Your Cooking Assistant</div>")
+    gr.Markdown("<div class='title'>Cookspert: Your Personal AI Chef</div>")
+
+    user_types = ["Professional", "Beginner", "Intermediate", "Expert"]
+
+    with gr.Column(scale=1, min_width=350):
+        # Prompt Section
+        prompt = gr.Textbox(label="Enter Your Cooking Question", placeholder="What would you like to ask?", lines=3)
 
-    user_types = ["Expert", "Intermediate", "Beginner", "Professional"]
+        # Model Selection Section
+        model_name = gr.Radio(label="Choose Model", choices=list(models.keys()), interactive=True)
 
-    with gr.Tabs():
-        with gr.TabItem("Ask a Cooking Question"):
-            with gr.Row():
-                with gr.Column(scale=2):
-                    prompt = gr.Textbox(label="Ask about any recipe", placeholder="Ask question related to cooking here...", lines=2)
-                    model_name = gr.Radio(label="Choose Model", choices=list(models.keys()), interactive=True)
-                    user_type = gr.Dropdown(label="User Type", choices=user_types, value="Beginner")
-                    submit_button = gr.Button("ChefGPT", elem_classes="button")
+        # User Type Selection
+        user_type = gr.Dropdown(label="Select Your Skill Level", choices=user_types, value="Home Cook")
 
-            response = gr.Textbox(
-                label="🍽️ Response",
-                placeholder="Your answer will appear here...",
-                lines=10,
-                interactive=False,
-                show_copy_button=True
-            )
+        # Submit Button
+        submit_button = gr.Button("chef gpt", elem_classes="button")
+
+        # Response Section
+        response = gr.Textbox(
+            label="Response",
+            placeholder="Your answer will appear here...",
+            lines=10,
+            interactive=False,
+            show_copy_button=True,
+            max_lines=12
+        )
 
-        submit_button.click(fn=process_input, inputs=[prompt, model_name, user_type], outputs=response)
+    submit_button.click(fn=process_input, inputs=[prompt, model_name, user_type], outputs=response)
 
 if __name__ == "__main__":
     demo.launch(server_name="0.0.0.0", share=True, debug=True)