15 days. I was shocked at how long it was going to take to upload a relatively small (20GB) image dataset from my local machine to a Google Cloud Platform (GCP) bucket.
In this article, I’ll go through the alternative approach that shortened this process to a 2-hour operation using a GCP Virtual Machine (VM). The key is creating a VM in the same zone as your bucket, downloading the file there and then transferring it to the bucket.
If you don’t have a bucket, go set that up first, and remember what zone you decide to use. …
I’m working through the fast.ai deep learning course this summer, and the instructor Jeremy often brags (in a good way, like a proud dad) about how students in the course can quickly get near-world-class models built with the fast.ai library. I wanted to put this idea to the test and decided to apply approaches from the first lecture only to a highly-cited cell paper from last year.
As a biologist, a Cell paper can be the highlight of one’s career. Cell publishes the best work from across all fields of biology, and the Kermany et. al. paper from last year…
I’m finishing up the material in the fast.ai Introduction to ML course and it has been fantastic. The learning philosophy, which the instructor Jeremy describes as a top-down approach, worked really well for my learning style. The general idea is that you learn HOW to do ML before you learn WHY it works — the same way you learn HOW to play baseball before learning WHY you might use one particular strategy at a given time.
Modeling biological networks and how drugs disrupt them | Altius Institute and UW | Climbing, running, skiing the PNW mountains