add gpu healthcheck

app.py

@@ -73,6 +73,7 @@ if IS_DEPLOYED and HF_CACHE_DIR:
     except (PermissionError, OSError):
         # If we can't create it, log but continue (might already exist from Dockerfile)
         pass
+        
 else:
     from dotenv import load_dotenv
     load_dotenv()
@@ -82,11 +83,32 @@ logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(
 logger = logging.getLogger(__name__)
 
 # Log environment setup for debugging
-logger.info(f"🌍 Environment: {'DEPLOYED' if IS_DEPLOYED else 'LOCAL'}")
 logger.info(f"📁 PROJECT_DIR: {PROJECT_DIR}")
-logger.info(f"📁 HuggingFace cache: {os.environ.get('HF_HOME', 'DEFAULT (not overridden)')}")
+logger.info(f"🌍 Environment: {'DEPLOYED' if IS_DEPLOYED else 'LOCAL'}")
 logger.info(f"🔧 OMP_NUM_THREADS: {os.environ.get('OMP_NUM_THREADS', 'NOT SET')}")
-    
+logger.info(f"📁 HuggingFace cache: {os.environ.get('HF_HOME', 'DEFAULT (not overridden)')}")
+
+
+import torch, sys
+try:
+    cuda_ = torch.cuda.is_available()
+    if gpu_check not in st.session_state:
+        st.write(f"GPU check skipped: {e.__str__}")  
+    print("CUDA:", cuda_)
+    logger.info("CUDA:", cuda_)
+    if cuda_:
+        if gpu_check not in st.session_state:
+            st.write(f"Device: {torch.cuda.get_device_name(0)}")
+        print("Device:", torch.cuda.get_device_name(0))
+        logger.info(f"Device: {torch.cuda.get_device_name(0)}")
+except Exception as e:
+    if gpu_check not in st.session_state:
+        st.write(f"GPU check skipped: {e.__str__}")
+    logger.error(f"GPU check skipped: {e.__str__}")
+    print("GPU check skipped:", e, file=sys.stderr)
+finally:
+    st.session_state.gpu_check = True
+            
 
 # Page config
 st.set_page_config(