# Kernel cell (piped to `colab exec`): waits for the detached ImageNet preparation (data5.py val,train) to finish, then # runs run_pilot_cell.py's logic (pilot.py skips the depth units already marked done on Drive). Keeps the kernel busy. import subprocess import time PREP_LOG = "/content/set5/logs/inet_prep.log" def prep_running(): return subprocess.run(["pgrep", "-f", "data5.py val,train"], capture_output=True).returncode == 0 while prep_running(): last = subprocess.run(f"tail -n 1 {PREP_LOG}", shell=True, capture_output=True, text=True).stdout.strip() print(time.strftime("%H:%M:%S"), "[prep]", last[:160], flush=True) time.sleep(60) print(time.strftime("%H:%M:%S"), "prep ended:", subprocess.run(f"tail -n 3 {PREP_LOG}", shell=True, capture_output=True, text=True).stdout, flush=True) exec(open("/content/set5/code/run_pilot_cell.py").read())