So how much do you pay in a month on average?
Can’t speak in months, I’m not a professional ML engineer (also, I have not encountered any fixed monthly costs yet in GCP, only pay-per-use).
But here’s an example from one week of the fastai course:
Preemptible N1 Predefined Instance Core running in EMEA, 19.82 Hours, 0.13 (€, n1-highmem-8 (8 vCPUs, 52 GB memory) )
Nvidia Tesla P100 GPU attached to preemptible VMs running in EMEA, 2.477 Hours, 0.98(€)
- 0.13 RAM and network
Total 1.24 € for ~ 20 hours of work
w/o deactivating the compute engine that would have been ~8 €.
Prediction for this month is ~ 13€ at the moment, with a much increased amount of usage this week. Of course it can be a bit a bit arduous restarting the VM when you need to activate the compute engine. Also, a pro will have another dev/running the model ratio.