Books I've read and highly recommend
Books I've read and highly recommend
Build a Large Language Model (From Scratch)
~by Sebastian Raschka, PhD
Currently reading
Very intuitive
Explained with code
~by Prof. Vijay Janapa Reddi (Harvard)
Currently reading
Key Points
Not a usual ML System textbook.
Deep dives at the hardware level to explain ML systems engineering.
Teaches why models fail at different levels of the stack
(Hardware → Runtime → Framework → Application)
Provides the foundation for Volume 2 of the book, which teaches scaling to multi-GPU training.
~by Harrison Kinsley & Daniel Kukieła
Key Points
Scratch Implementation of Deep Learning Concepts.
Easy to follow and understand.
Mathematics well explained.
Atomic Habits
Some ideas from the book.
1% Rule: Improving by 1% daily can lead to being 37 times better in a year.
The Plateau Effect: Small habits compound over time, even if initial results seem invisible
Zero to One
(by PayPal founder, an early investor in SpaceX, FaceBook)
Highlights
All startup advices are still relevant after a decade.
0 to 1 means innovations (the hard part) discussed in detail, while 1 to n is globalization (the easy part).
Excellent Advice for Living
~presented as a list of 406 pieces of advices rather than in paragraphs.
Here are a few of my favorites:
• Never hesitate to invest in yourself—to pay for a class, a course, a new skill. These modest expenditures pay outsized dividends.
• When you find yourself procrastinating, don’t resist. Instead lean into it. Procrastinate 100%. Try to do absolutely nothing for 5 minutes. Make it your job. You’ll fail. After 5 minutes, you’ll be ready and eager to work.
• Think in terms of decades, and act in terms of days.