I built GPT-1 from scratch and trained it to perform two downstream tasks. (To be more precise, I developed this version of GPT-1 in collaboration with Claude, based upon a transformer_tf.py script that I had created from scratch.)

A detailed explanation of the internal architecture of GPT-1 will be published in The Engineer’s Guide To Deep Learning in the near future (within a few years).

In this post, I will only briefly demonstrate how to use it. Please note that this is an alpha version, and many refactoring updates are planned for the future.

This is a blog that combined and re-edited version of the two related blogs: The 3 Steps to a Faster EXPLAIN ANALYZE and Two More Steps to a Faster EXPLAIN ANALYZE.

I recalled performing model verification for pgpool-II (a PostgreSQL replication tool) using SPIN in 2010, so I’m sharing some of the results.

In previous posts, post1 and post2, I demonstrated that it’s possible to continuously count the actual rows from EXPLAIN ANALYZE with an overhead of about 2.0%.

The next step is to examine methods for correcting the selectivity (or cardinality) estimation results.

Based on the improvements shown in this blog post, I will show the results of further optimizations.

In this blog, I’ll challenge conventional wisdom and explore methods to reduce overhead as much as possible. To that end, I will first discuss data types before explaining the specific improvements.

1-Line Abstract: Cut the overhead of EXPLAIN ANALYZE by roughly 77 [%].

I show a tip of autovacuum management.

As frequently mentioned, the table bloat problem is one of the most annoying things in managing PostgreSQL. Several things can cause that problem, and Autovacuum is one of them.

On March 8th, Datadog totally paralyzed around the world and it continued a few days.

I was interested in the caurses of that failure from the point of view of my pure engineering interest. However, Datadog have not disclosed anything yet.

I list up three kinds of technologies I am interested in except computing. If I were young, I would try to dive into these fields.

I’ve just released pg_plan_inspector that is a framework to monitor and improve the performance of PostgreSQL using Machine Learning methods.

One movie is worth a thousand words. Watch this movie.