Update app.py

app.py

@@ -2,8 +2,31 @@ import gradio as gr
 import os
 import subprocess
 from pathlib import Path
+import time
+import requests
+
+def check_services():
+    """Check if all required services are running"""
+    services = [
+        ("Controller", "http://localhost:21001"),
+        ("API Server", "http://localhost:8000"),
+        ("Model Worker", "http://localhost:8080")
+    ]
+    
+    for service_name, url in services:
+        try:
+            requests.get(url)
+            print(f"{service_name} is running")
+        except requests.exceptions.ConnectionError:
+            return False, f"{service_name} is not running"
+    return True, "All services are running"
 
 def check_training_status():
+    # First check if services are running
+    services_ok, message = check_services()
+    if not services_ok:
+        return message
+    
     results_dir = Path("/app/results")
     if not results_dir.exists():
         return "Training hasn't started yet."
@@ -12,6 +35,11 @@ def check_training_status():
     return f"Completed {iterations} training iterations."
 
 def start_training(model_path, instruct_count, max_iter):
+    # Check if services are running
+    services_ok, message = check_services()
+    if not services_ok:
+        return message
+    
     os.environ["MODEL_PATH"] = model_path
     os.environ["INSTRUCT_COUNT"] = str(instruct_count)
     os.environ["MAX_ITER"] = str(max_iter)
@@ -70,4 +98,14 @@ with gr.Blocks() as iface:
     )
 
 if __name__ == "__main__":
+    # Wait for services to be ready
+    print("Waiting for services to start...")
+    while True:
+        services_ok, message = check_services()
+        if services_ok:
+            break
+        print(message)
+        time.sleep(5)
+    
+    print("All services are running, starting web interface...")
     iface.launch(server_name="0.0.0.0", server_port=7860)