success

runners/tensorboard_runner.py

@@ -0,0 +1,37 @@
+import os
+import subprocess
+import sys
+
+def find_free_port(start_port):
+    import socket
+    port = start_port
+    while True:
+        with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+            if s.connect_ex(('localhost', port)) != 0:
+                return port
+            port += 1
+def run(exp_name, exp_root="experiments",port=None):
+    port = 6007 if port is None else port
+    max_attempts = 10
+    project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+    tensorboard_root = os.path.join(project_root, exp_root, exp_name, "tensorboard")
+    
+    for attempt in range(max_attempts):
+        try:
+            print(f"Trying to launch TensorBoard on port {port}...")
+            subprocess.check_call([
+                sys.executable, "-m", "tensorboard.main",
+                f"--logdir={tensorboard_root}",
+                f"--port={port}"
+            ])
+            break
+        except subprocess.CalledProcessError as e:
+            print(f"Port {port} is in use, trying next port...")
+            port = find_free_port(port + 1)
+    else:
+        print("Failed to launch TensorBoard after multiple attempts.")
+
+if __name__ == "__main__":
+    exp_root = "experiments"
+    exp_name = "sample_train_100_item_overfit_foreground_0"
+    run(exp_name,exp_root,port=6009)