Stop button: write flag file, wait 120s for clean save, fallback terminate

app.py

@@ -1,5 +1,5 @@
 import gradio as gr
-import subprocess, sys, os, threading, json, datetime
+import subprocess, sys, os, threading, json, datetime, time
 import matplotlib
 matplotlib.use('Agg')
 import matplotlib.pyplot as plt
@@ -72,6 +72,17 @@ def stop_training():
     with _lock:
         if not _proc or _proc.poll() is not None:
             return "No training process is currently running.", *build_plots()
+        # Write flag file — train.py checks this every step and saves weights before exiting
+        with open('/app/stop_requested', 'w') as f:
+            f.write('stop')
+
+    # Wait up to 120s for train.py to save weights and exit cleanly
+    for _ in range(24):
+        time.sleep(5)
+        if _proc.poll() is not None:
+            break
+    else:
+        # Fallback: hard terminate if train.py didn't respond
         _proc.terminate()
 
     m = _load_metrics() or {}