Upload runner.py with huggingface_hub

runner.py

@@ -2,34 +2,54 @@
 import os
 import subprocess
 import time
+import sys
+import threading
+from http.server import HTTPServer, SimpleHTTPRequestHandler
 from huggingface_hub import HfApi
 
+# Start a dummy HTTP server on 7860 so Hugging Face marks the space as 'Healthy' and doesn't kill it.
+def start_health_server():
+    class HealthCheckHandler(SimpleHTTPRequestHandler):
+        def do_GET(self):
+            self.send_response(200)
+            self.end_headers()
+            self.wfile.write(b"Training in progress...")
+    server = HTTPServer(('0.0.0.0', 7860), HealthCheckHandler)
+    server.serve_forever()
+
+threading.Thread(target=start_health_server, daemon=True).start()
+
+def log(msg):
+    print(msg, flush=True)
+
 def main():
     repo_id = "SParsh003/LifeOS-Trainer"
     api = HfApi(token=os.environ.get("HF_TOKEN"))
     
+    log("===== STARTING TRAINING RUNNER =====")
+    
     # Verify GPU exists
     try:
         subprocess.run(["nvidia-smi"], check=True)
-        print("GPU detected. Proceeding with training setup...")
+        log("GPU detected. Proceeding with training setup...")
     except:
-        print("No GPU detected. Idling to prevent loops.")
+        log("No GPU detected. Idling to prevent loops.")
         while True: time.sleep(3600)
         
     if os.path.exists("training_done.txt"):
-        print("Training already completed. Idling.")
+        log("Training already completed. Idling.")
         while True: time.sleep(3600)
         
-    print("Cloning repository...")
+    log("Cloning repository...")
     os.system("git clone https://huggingface.co/spaces/SParsh003/LifeOS-Personal-Chaos-Agen LifeOS")
     os.chdir("LifeOS")
     
-    print("Installing Unsloth, TRL, and dependencies (this takes a few minutes)...")
+    log("Installing Unsloth, TRL, and dependencies (this takes a few minutes)...")
     os.system("pip install unsloth")
     os.system("pip install --no-deps trl peft accelerate bitsandbytes datasets transformers sentencepiece")
     os.system("pip install matplotlib pydantic fastapi requests")
     
-    print("Starting GRPO Training (50 episodes)...")
+    log("Starting GRPO Training (50 episodes)...")
     cmd = [
         "python", "-m", "lifeos.training.train_grpo",
         "--real-gpu",
@@ -41,17 +61,17 @@ def main():
     # Run and stream output
     process = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, universal_newlines=True)
     for line in process.stdout:
-        print(line, end="")
+        print(line, end="", flush=True)
     process.wait()
     
-    print("Training complete!")
+    log("Training complete!")
     os.chdir("..")
     with open("training_done.txt", "w") as f:
         f.write("done")
         
-    print("DOWNGRADING HARDWARE TO CPU TO STOP BILLING...")
+    log("DOWNGRADING HARDWARE TO CPU TO STOP BILLING...")
     api.request_space_hardware(repo_id=repo_id, hardware="cpu-basic")
-    print("Hardware downgraded successfully.")
+    log("Hardware downgraded successfully.")
     
     while True:
         time.sleep(3600)