ameyab retweetledi
You can't write evals for what you haven't seen yet.
We talked with @braintrust VP Field CTO, Ameya Bhatawdekar, at HumanX about why production observability is what makes evals actually work.
English
ameyab
213 posts
ameyab
@ameyab
now: @braintrust; was: AI @ Dropbox, AI at @Office, AI @facebookai; built SmartScreen
Seattle Katılım Nisan 2008
205 Takip Edilen64 Takipçiler
Can we compress large language models for better perf?
"SparseGPT: Massive Language Models can be Accurately Pruned in One Shot"
Eliminates the need to use/store 50% of weights for a 175B param model with no significant sacrifice in perf
arxiv.org/pdf/2301.00774…
Here's how 👇
English
My first year, as an engineering manager, I’d try to collect as much data I could about what people did. Spent ~6-8 hrs per person writing perf reviews, trying to make sure I didn’t miss a thing, working through weekends.
It was exhausting. Then I made a change (cont’d):
Gergely Orosz@GergelyOrosz
Performance reviews are coming up. An objective way to summarize your own achievements is to use numbers. Those numbers are facts, hard to argue with, and can be compared, should anyone want to do so. Some ideas for numbers. What else have you seen used in perf reviews?
English
Here are all the ways to get around ChatGPT's safeguards:
[1/n]
English
ChatGPT has crossed 1M+ users in just 5 days.
To compare, it took Netflix 41 months, FB - 10 months, and Instagram - 2.5 months.
But many haven’t yet realized its full potential.
Here are the 10 mindblowing things you can do using it right now:
English
With the ChatGPT wave rolling over twitter I also want to celebrate that the TRL library crossed 500 stars⭐️🎉
If you want to play with reinforcement learning for language models (the tech behind ChatGPT) yourself checkout the repo&library:
github.com/lvwerra/trl
English
ChatGPT is blowing up the internet.
Here is a thread showing some interesting examples I found. 🧵
English
Did you know, that you can build a virtual machine inside ChatGPT? And that you can use this machine to create files, program and even browse the internet? engraved.blog/building-a-vir…
English
How to retain everything you learn.
The Spaced Repetition Method (science-backed):
English
Every Founder in Year 1:
"I love my company. These people are my family and I will run this business forever."
Every Founder in Year 8:
"I hate my life and my employees hate me. Please buy my business and let me go away for a very long time."
English
The new slogan of AI might as well be "show, don't tell"!
Finding the BLOOM paper was really hard. Searching Google & multiple blogs, all I found were demos.
If someone's still interested in papers: "BLOOM: A 176B-Parameter...Language Model" arxiv.org/abs/2211.05100
English
🐙TorchScale - A Library for Transformers at (Any) Scale
TorchScale is a PyTorch library to scale up Transformers efficiently and effectively. Something worth checking out if you are working on scaling Transformer models.
github.com/microsoft/torc…
English
Some user notes on V2! 🧵
V2 prompts different and will take a while for folk to get used to.
V2 is trained on two models, a generator model and a image-to-text model (CLIP).
V1 used the CLIP-L14 model open sourced by @OpenAI which was great, but nobody knew what was in it 1/
Stability AI@StabilityAI
We are excited to announce the release of Stable Diffusion Version 2! Stable Diffusion V1 changed the nature of open source AI & spawned hundreds of other innovations all over the world. We hope V2 also provides many new possibilities! Link → stability.ai/blog/stable-di…
English
🔴 zety.com
🔴 resume.com
🔴 flowcv.com
🔴 resume.io
🔴 resumegenius.com
🔴 resumebaker.com
🔴. enhancv.com
Debugging your PyTorch code today?
Here's a lovely, little open-source project for a lovely Sunday!
Lovely tensors -- tensors, ready for human consumption:
github.com/xl0/lovely-ten…
English
Google, Microsoft, Adobe, IBM, Palo Alto Networks, and now Twitter run by CEOs who grew up in India. Wonderful to watch the amazing success of Indians in the technology world and a good reminder of the opportunity America offers to immigrants. 🇮🇳🇺🇸 (Congrats, @paraga!)
English
I am putting together a list of platforms that you can explore if you are looking for economical AI/ML training and inference -
1. banana .dev - Ship ML to Prod, instantly. Scalable inference hosting for your machine learning models on serverless GPUs.
Thread ->
English
Today @pmarca was asked by @tylercowen to explain a Web3 use case.
I clipped this gem from 28:08 of Conversations With Tyler.
Highly recommended...
English