🧠 ML training
Keep your Mac awake for an overnight model training run
You kick off a multi-hour fine-tune before bed and want it finished by morning, not frozen at epoch 3. AwakeMate keeps your Mac awake while the run works.
Free 14-day trial · one-time $39.99 · no telemetry · macOS 13+
The workflow
You start a local fine-tune or training run in Python - PyTorch or Apple's MLX - on your Apple Silicon Mac. It runs for hours, hammering the GPU and CPU, writing checkpoints as it goes, and you walk away expecting it to be done by morning.
What happens when the Mac sleeps
When macOS hits idle sleep, your training process is frozen mid-step. It does not auto-resume when you wake the Mac, so a multi-hour run interrupts and may never complete. Worse, a checkpoint that was mid-write can be left half-written, leaving you with a corrupt file and lost progress.
How AwakeMate keeps it alive
AwakeMate detects your training process and holds the Mac awake while it runs. Fine-tunes launched via Python usually show up as python or python3; with Deep CLI detection (on by default) it reads the real script name instead. Leave Sleep when tools go idle OFF here - a quiet checkpoint-writing lull looks idle but the run is not done, and you do not want the Mac napping through it.
- Open AwakeMate's Apps pane and add your Python interpreter (or let auto mode pick it up) so it keeps the Mac awake while the run is active.
- Confirm Deep CLI detection is on so a
pythonprocess resolves to your actual training script. - Leave Sleep when tools go idle OFF, start the run, and check the menu-bar item shows the Mac is being kept awake.
Tip: Watch the first checkpoint write, then trust it. If you have set quiet hours or a low-battery sleep rule, remember those can still force sleep - keep the Mac on mains power for an overnight run.
Tools in this workflow
Frequently asked
Why not use "sleep when tools go idle" for training?
Training writes checkpoints in bursts, with quieter I/O lulls in between. Idle-sleep could mistake one of those lulls for the job finishing and let the Mac sleep, freezing the run. Leave it off for training.
My training process just shows as "python" - will AwakeMate still match it?
Yes. With Deep CLI detection on (the default) it reads the real script name behind a generic python process, so you can target your training script specifically.
Does AwakeMate change my energy settings or speed up training?
No. It holds the same IOKit power assertion as caffeinate, preventing idle sleep only. It never touches your saved energy settings and does not affect training speed.
Will this keep running with the lid closed, even on battery?
Yes. Turn on "Keep awake with the lid closed" in Settings - Options and approve the one-time helper; AwakeMate keeps the Mac awake with the lid shut while the run is active, and dims the built-in screen so it is not draining the battery under the lid. It works on most Apple Silicon Macs (Apple forces sleep on some, so close your lid once to confirm), and on battery too - though a long run drains faster, so keep it on power. See it in action.