Factory-AI for Deep Learning Purposes

Taught how to effectively use the internal HPC cluster to optimize Deep Learning experiments.

The course included:

• SLURM Tutorial:
  • Principles of nodes, jobs, submissions, queue
  • How to submit a job
    • configuration to maximize usage of partition resources
    • multi-node and multi-process setting
• Pytorch Tutorial:
  • Dataloader
  • Distributed training
    • along with SLURM by using set environment variables
  • Avoid CPU/GPU synchronization

After this course, a noticable number of experiments were achieved faster on the cluster and computational resources were better used.