Amol

LLM Perf @ E2E Networks
In the past: founder hyperML (compute platform)

[Amol @aumbark]

@MatijaSosic can take the model selection and routing to next level

[Amol @aumbark]

@mitsuhiko @thdxr I think it's effective routing. An efficient prefix tree that scores workers and it's distance to storage. Some of nvidia dynamo components offer good glimpse of this + dpu on nodes help. you can get closer ~95% . But 99% at this scale is unbelievable.

[Amol @aumbark]

@nileshtrivedi spot on. This prob follows kahnman theory.

Reading a book requires System II and it is a finite resource. Most books are deeper in context and one has to engage at that level.

Visual media and short tweets dont demand thinking. and hence more preferable to general population.

[Amol @aumbark]

@CoachDanGo Do mention appropriate dose.

There are many in the market that are 2000mg - a nuke for bowel movements :)

[Amol @aumbark]

@CoachDanGo definitely a vampire bit him.

[Amol @aumbark]

@hubermanlab most of the body metrics can be looked at like hrv, bp. its a period of time that matters, not one instance

[Amol @aumbark]

@hubermanlab I think this should be looked at for a period of time (5-6 days no alarm?) than a particular day.

[Amol @aumbark]

@snj_1970 Sometimes it never did. And when it did, the stress in rats was less after the shock than during the music.

The news about third wave and covid seems to be playing a similar situation in our head.

[Amol @aumbark]

@snj_1970 I remember reading about an experiment where lab rats were given shocks and a music was played in the background at the same time.
Overtime, the rats learned to recognise the music and got scared just at the sound of it, waiting for the shock to arrive.

[Amol @aumbark]

@PrasoonPratham It would be awesome to see a journey of building a model from scratch to 'live' usage.

[Hashnode @hashnode]

A #Christmas gift from the Hashnode community to the #devcommunity on Twitter. 🙌

Here's our first FREE community-contributed book 📕 "Become A Better Dev In 2021"

40 helpful guides written by 30⭐️ awesome writers from @hashnode 🧑‍💻 👩‍💻

Download ⬇️
https://t.co/ezfVyWG3KA

[Amol @aumbark]

This is an awesome roadmap to every useful resource on Deep Learning https://t.co/ao1PBJ4yfy

[Amol @aumbark]

@PrasoonPratham Awesome problem to start with.

Check out quantopian in case you haven't.

Also do try on bit-coin data. Less variables compared to stocks.

You might end up making 💰 :)

[Amol @aumbark]

In this thread we covered:
- Why use Gradient Descent
- What model training has in common with Bubble Shooter and Tetris (mobile games)
- When to stop training a model
- What the heck is Learning Rate

#100DaysOfCode #100DaysOfMLCode #MachineLearning #DEVCommunity

[Amol @aumbark]

✍️ A super simple Intro to Gradient Descent in Machine Learning

In this thread we will learn:
- Why use Gradient Descent
- What model training has in common with Bubble Shooter and Tetris (mobile games)
- When to stop training a model
- What the heck is Learning Rate
🧵 👇

[Amol @aumbark]

In future posts - Stochastic Gradient Descent, Adaptive Learning Rate, and more!
🧵☝️