When: 17.45 (Talks begin at 18:00) Agenda: 1) Martin Omander (Google) - Serverless Computing; 2) John Angelmo - Terraform and GCP; More eyes on your infrastructure and letting each team handle their own infrastructure. 3) Leonid Kuligin (Google) - Training your ML Model with Google Cloud Platform.
Drinks: Yes
Food: Yes
Coming home on time: Nope, this is a full on night.
If you want to spend more time building your product and less time managing servers, you should consider serverless computing. In this talk we will consider common use cases and how they can be implemented quickly using serverless products from Google.
1. benefits you get with training your models with the Cloud, basic options to do that
2. main considerations when using a rent-a-VM option for model training. CPUs vs GPUs vs TPUs.
3. vertical vs. horizontal training scalability. distributed model training: data vs model parallelism.
4. managed services for model training offered by GCP