You are migrating a CPU-intensive machine learning application to GCP and need to optimize costs. Your application can tolerate some downtime. What type of VMs and scheduling should you use?

*Explanation: _Preemptible VMs with Custom Machine Scheduling._ A. On-demand scheduling would make preemptible VMs costly. B. Custom VMs with a high number of vCPUs could be expensive even with discounts. C. N2D High-CPU VMs are expensive for on-demand use. D) E2 VMs are not optimized for CPU-intensive tasks. E) Preemptible VMs with Custom Machine Scheduling optimize both cost and resources. F) N2 VMs are costly even with discounts.*