divyanshsinghvi's profile picture. Nobody yet!

Divyansh

@divyanshsinghvi

Nobody yet!

If you’re starting out with transformers, after reading about it. A great next step: build something like a smaller GPT-2 decoder-only transformer from scratch. This walkthrough makes it very doable and really helped clear up my own understanding: …ter1-transformer-interp.streamlit.app

arena-chapter1-transformer-interp.streamlit.app

Chapter 1 Transformer Interpretability

This GitHub repo hosts the exercises and Streamlit pages for the ARENA program. (Note that the na...


Has @IndiGo6E airlines became too prudent? 1. Too many flight delays. 2. The staff too arrogant to take note of allergic reaction of food items or take any step to resolve your query? 3. No mention of contents in their meal of the day. @MoCA_GoI @RamMNK


Divyansh 님이 재게시함

One of the most useful tricks I learnt from @giffmana: automatically starting pdb debugger when program crashes. It's just 5 lines of code that has probably saved me countless hours in trying to debug tricky bugs.

_gaganm's tweet image. One of the most useful tricks I learnt from @giffmana: automatically starting pdb debugger when program crashes.

It's just 5 lines of code that has probably saved me countless hours in trying to debug tricky bugs.

First time I have seen a programming language trending, and it's not the obvious answers : C++, Python instead OCaml. Interesting times! I wonder how good the recent version update is?


Pretty good summarization of using RL for LM vs Instruct tuning(supervised) for LM.

I was puzzled for a while as to why we need RL for LM training, rather than just using supervised instruct tuning. I now have a convincing argument, which is also reflected in a recent talk by @johnschulman2 . I summarize it in this post: gist.github.com/yoavg/6bff0fec…



Don't know which India author is talking about?


Divyansh 님이 재게시함

When the curtains come down. After keeping all of you updated with Covid19 data for the last 16 months, we will be stopping our operations from 31st October, 2021. Why are we doing this? What will happen next? This blog post has all the details 👇 blog.covid19india.org/2021/08/07/end/


Amazing!

Parkour is the perfect sandbox for the Atlas team to experiment with new behaviors. Through jumps, balance beams, and vaults, we push Atlas to its limits to discover the next generation of mobility, perception, and athletic intelligence. bit.ly/3iTIxco



Loading...

Something went wrong.


Something went wrong.