Grok-1 support in gpt-fast at faster(?) than anyone else has reported so far. 75 tok/s for a 300B+ parameter model on an 8xA100 node.

If I understand correctly, ColossalAI reported 15 seconds to generate 100 tokens. gpt-fast takes 4.2 seconds to generate *400* tokens.

This is a very disturbing story. 1 in 10 people who acquired COVID in Victorian hospitals died. The hospitals refuse to talk about it. What is the point of a hospital if it makes patients sick? abc.net.au/news/2024-05-0…

Throwback to this amazing moment exactly five years ago (!!!) with Jeremy Howard & Jason Antic at F8 where we presented our work on DeCrappification fast.ai/posts/2019-05-…

Short overviews of writings on consciousness:

What is it Like to be a Bat? : Nagle - qualia is a thing
The Conscious Mind : Chalmers - science won't explain consciousness
Consciousness Explained : Dennett - consciousness is not a thing, or a natural brain function.
Shadows of…

Coming soon to Gradio.

Can anyone guess what gr.render() does?

I made 700 coding tutorials on ChrisAlbon.com. 5-6 years later maybe 20% of them were so outdated they wouldn’t run.

these models really need to be evaluated with something like RULER (github.com/hsiehjackson/R…), needle-in-a-haystack is not sufficient for me to believe that performance doesn't suffer across the long-context window

I’m excited to share that I’ve joined Answer.AI as member of the technical staff for R&D!

I’m really excited to join such a small yet highly creative and productive team—led by Jeremy Howard and Eric Ries—in pursuit of practical, development-driven LLM breakthroughs!

a thread with some of my top animations 📍🧵
twitter.com/matthen2/statu…

to tell if a maze is solvable, just hang it by its corners! The first maze stays in one piece, so there is no path from the entrance at the top to the exit at the bottom. The second maze splits apart along the solution.

its true. we are working on it :)

Matthew Carrigan Try WizardLM’s WizardLM-2 8x22B, beats every open source LLM I’ve tried, including Llama-3 70B Instruct, by far

Introducing ChatQA-1.5, a family of models that surpasses GPT-4-0613 and Command-R-Plus on RAG and conversational QA.

ChatQA-1.5 has two variants:
Llama3-ChatQA-1.5-8B, huggingface.co/nvidia/Llama3-…
Llama3-ChatQA-1.5-70B, huggingface.co/nvidia/Llama3-…

We also open source our instruction…

Prometheus 2: An Open Source Language Model Specialized in Evaluating Other Language Models

Presents a more powerful evaluator LM than its predecessor that closely mirrors human and GPT-4 judgements

repo: github.com/prometheus-eva…
abs: arxiv.org/abs/2405.01535

Thanks for highlighting Ark Biotech, Sen Fetterman! It's great to see you helping your constituents envision a future where meat production is as clean and cruelty free as brewing beer, instead of looking like a slaughterhouse floor. 🇺🇸🥩

At least it's confirmed that no, it's not a 'Large Action Model' - it's an LLM.. lol

Friday afternoon? It's not closing time yet Gradient

Llama-3-70b up to 262k context length with our improved dataset for training.

We need to keep up with Wing Lian (caseus) and Eric Hartford

My YouTube viewership across two channels is 99.7% male... I am probably (almost) all of the 0.3% female views on my channels.

If anyone has tips to improve this ratio, lemme know lol

This is one of the weirder instincts I feel after over a decade in startups

Crisp well-structured plans are highly convincing but indicate that we are engaging with a fantasy “Lego brick” version of reality

Almost all the “head against wall” moments I’ve had, and that I’ve seen…

# CUDA/C++ origins of Deep Learning

Fun fact many people might have heard about the ImageNet / AlexNet moment of 2012, and the deep learning revolution it started.
en.wikipedia.org/wiki/AlexNet

What's maybe a bit less known is that the code backing this winning submission to the…