Update app.py

app.py

@@ -6,6 +6,8 @@ import gradio as gr
 import os
 import logging
 from unsloth import FastLanguageModel
+import subprocess
+
 logging.basicConfig(
     level=logging.DEBUG,  # Set the logging level to DEBUG to capture all messages
     format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
@@ -72,11 +74,22 @@ Category List : ["Dairy & Eggs", "Beverages & Snacks", "Cleaning & Hygiene", "Gr
 '''
 @spaces.GPU()
 def chunk_it(inventory_list, user_input_text):
-    num_elements = (5 * 1024 * 1024) // 4
-    tensor = torch.randn(num_elements, dtype=torch.float32)
-    tensor_gpu = tensor.to('cuda')
-    logger.info("Loading model and tokenizer...")
+    # Check for CUDA and NVIDIA-related errors
     try:
+        # Check for GPU devices
+        device_count = torch.cuda.device_count()
+        logger.info(f"Number of GPU devices: {device_count}")
+        if device_count == 0:
+            raise RuntimeError("No GPU devices found.")  # Raise an error if no GPUs are detected
+
+        # Check CUDA version using subprocess
+        process = subprocess.run(['nvcc', '--version'], capture_output=True, text=True)
+        cuda_version = process.stdout.strip()
+        logger.info(f"CUDA version: {cuda_version}")
+        if 'not found' in cuda_version.lower():
+            raise RuntimeError("CUDA not found.")  # Raise an error if CUDA is not found
+
+        # Load model and tokenizer (your original code)
         model, tokenizer = FastLanguageModel.from_pretrained(
             model_name = "VanguardAI/CoT_multi_llama_LoRA_4bit",
             max_seq_length = 2048,