Obsidian Source: Drafts / Tech Blog
Summary
Pending synthesis from local Obsidian source.
Original source title: Tech Blog
Extracted Preview
Why is JAX so awesome(read tough)?
- Working on Autodiadax, to understand the inner core of JAX.
- Interpreters, Evaluators, Tree, Flat functions, it all gets confusing after some time.
- Used NotebookLM to generate a podcast to understand how the flow is, just to be confused by the terminologies.
- I was able to debug through error stack, so things are going OK.
- Lol Giffmana and Many people reached out to help. Will go through it.
Is the answer to universe JAX?
JAX: Why is Everyone So Excited About This Framework?
Unlocking the machine FLOPS is probably the main goal of any library. JAX does it, in the way that it allows researchers to write Python programs that are automatically scaled to leverage accelerators and supercomputers.
What is meant by pure function?
A function is "pure" when it does not have any side effects, statistically composed(relative to a set of primitive functions, when it can be represented as a set of static data dependency graph on those primitives.). PSC sub-routines.
Francois Chollet spoke recently that almost all top players in Generative AI are pivoting to JAX as it is fast, scales really well(has good distributed support), and TPU support is there as well. JAX does the heavy lifting for you(the compiler optimizes your code), all you have to do is to define what you want(basically computation graphs of inputs and outputs)
Integration Notes
- Source folder:
/home/yashs/Documents/Docs/Obsidian/Research-Notes - Local source:
/home/yashs/Documents/Docs/Obsidian/Research-Notes/Drafts/Tech Blog.md - Raw copy:
raw/obsidian/research-notes/Drafts/Tech Blog.md