update space

app.py

@@ -7,6 +7,7 @@ import numpy as np
 from huggingface_hub import login
 from pymongo import MongoClient
 from transformers import AutoModelForCausalLM, AutoTokenizer
+from concurrent.futures import ThreadPoolExecutor
 
 # Ensure the Hugging Face token is set
 tok = os.environ.get('HF_Token')
@@ -32,11 +33,18 @@ print("TensorFlow GPU Available:", tf.config.list_physical_devices('GPU'))
 
 # Set TensorFlow to use mixed precision with available GPU
 from tensorflow.keras import mixed_precision
+
 if len(tf.config.list_physical_devices('GPU')) > 0:
-    # Set mixed precision policy to use float16 for better performance on GPU
-    policy = mixed_precision.Policy('mixed_float16')
-    mixed_precision.set_global_policy(policy)
-    print("Using mixed precision with GPU")
+    # Ensure the GPU supports mixed precision
+    gpu_device = tf.config.list_physical_devices('GPU')[0]
+    gpu_info = tf.config.experimental.get_device_details(gpu_device)
+    if 'compute_capability' in gpu_info and gpu_info['compute_capability'][0] >= 7:
+        # Set mixed precision policy to use float16 for better performance on supported GPUs
+        policy = mixed_precision.Policy('mixed_float16')
+        mixed_precision.set_global_policy(policy)
+        print("Using mixed precision with GPU")
+    else:
+        print("GPU does not support mixed precision or may not provide significant benefits. Using default precision.")
 else:
     print("Using CPU without mixed precision")
 
@@ -46,11 +54,16 @@ try:
     device_name = '/GPU:0' if len(tf.config.list_physical_devices('GPU')) > 0 else '/CPU:0'
     print(f"Loading models on {device_name}...")
     with tf.device(device_name):
-        # Load the poultry disease detection model
-        my_model = load_model('models/disease_model.h5', compile=True)
+        def load_models():
+            with ThreadPoolExecutor() as executor:
+                future_disease_model = executor.submit(load_model, 'models/disease_model.h5', compile=True)
+                future_auth_model = executor.submit(load_model, 'models/auth_model.h5', compile=True)
+                return future_disease_model.result(), future_auth_model.result()
+
+
+        # Load models concurrently
+        my_model, auth_model = load_models()
         print("Disease detection model loaded successfully.")
-        # Load the authentication model
-        auth_model = load_model('models/auth_model.h5', compile=True)
         print("Authentication model loaded successfully.")
         print(f"Models loaded successfully on {device_name}.")
 except Exception as e:
@@ -66,6 +79,7 @@ recommend = {
     3: 'Ponston'
 }
 
+
 class PoultryFarmBot:
     def __init__(self):
         self.db = db  # MongoDB database for future use
@@ -114,7 +128,7 @@ class PoultryFarmBot:
     # Generate a detailed response using Llama 2 for disease information and recommendations
     def generate_disease_response(self, disease_name, status, recommendation):
         print("Generating detailed disease response...")
-        # Create a prompt for Llama 3 to generate detailed disease information
+        # Create a prompt for Llama 2 to generate detailed disease information
         prompt = (
             f"The disease detected is {disease_name}, classified as {status}. "
             f"Recommended action: {recommendation}. "
@@ -136,6 +150,7 @@ class PoultryFarmBot:
         print("Invalid image provided.")
         return "Please provide an image of poultry fecal matter for disease detection.", None, None, None
 
+
 # Initialize the bot instance
 print("Initializing PoultryFarmBot instance...")
 bot = PoultryFarmBot()
@@ -154,7 +169,8 @@ if tokenizer.pad_token is None:
     model.resize_token_embeddings(len(tokenizer))
     print("Pad token added and model resized.")
 
-# Define Llama 2 response generation
+
+# Define Llama 3.2 response generation
 def llama2_response(user_input):
     try:
         print("Generating response using Llama 2...")
@@ -178,6 +194,7 @@ def llama2_response(user_input):
         print(f"Error generating response: {e}")
         return f"Error generating response: {str(e)}"
 
+
 # Main chatbot function: handles both generative AI and disease detection
 def chatbot_response(image, text):
     print("Received user input for chatbot response...")
@@ -193,9 +210,10 @@ def chatbot_response(image, text):
             return diagnosis  # Return only the diagnostic message if no disease found
     else:
         print("No image provided, using Llama 3.2 for text response...")
-        # Use Llama 2 for more accurate responses to user text queries
+        # Use Llama 3.2 for more accurate responses to user text queries
         return llama2_response(text)
 
+
 # Gradio interface styling and layout with ChatGPT-like theme
 print("Setting up Gradio interface...")
 with gr.Blocks(theme=gr.themes.Soft(primary_hue="green", neutral_hue="slate")) as chatbot_interface: