Kornesh Kanan

SIREN: Implicit Neural Representations with Sine

Machine learning is just a function takes encoded representation of input $x$ and maps it to encoded representation of output $y$. Does the way we choose to encode the input has an impact on the model? Can we represent an image better than a

Kubernetes

Setting up nginx-ingress without Google Load Balancer in GKE

It's nice to have a load balancer to dynamically distribute incoming traffic across nodes inside Kubernetes cluster. However, it is a bit of overkill when you are running your side projects on GKE with a single node and it costs additional ~$20 a month

DevBytes

Programmatically accessing Google Sheets via Service Account

I was working on automation that involves reading Google Sheets from server side and then writing some values back. Initially this seemed straight forward but led me to waste a couple of hours pursuing the false start of "Domain Wide Delegation" which would allow

Kubernetes

On GKE Cluster Management Fee

On June 6, 2020, your Google Kubernetes Engine (GKE) clusters will start accruing a management fee, with an exemption for all Anthos GKE clusters and one free zonal cluster. ...I was utterly disappointed when I received this email last night. It introduces additional $75

Machine Learning

10x faster RoBERTa tokenizer with Custom Tokens support

For my Question Answering Kaggle competition, I wanted to experiment replacing the BERT model with RoBERTa. This means I need to reencode and retokenize the entire Natural Questions dataset into TFRecords. This process was already taking hours with the WordPiece tokenizer used for the

Kubernetes

Supporting clients that does not support SNI in EKS

SNI is something that's been enabled by default for most modern browsers and http clients. It allows us to serve multiple different SSL certs on the same IP address and TCP port. This is incredibly useful for multi tenancy in EKS when used along

Quantum

Understanding Computational Model of Quantum Computers

When I look at computers I would at least have some intuition how it all works in theory. There was bunch of theories we had to learn in university to reason with the mathematical models of it's computation. However, I had no idea theoretically

DevBytes

Improving slow EFS throughput when copying thousands of small files

We were in the process of migrating our customers from a private cloud to AWS EKS. During our dry-runs, we noticed Customer A with 21GB worth of files had it transferred in 32 minutes from a local disk in the private cloud to EFS.