AI - AI Dynamics

Poe Enables Fine-Tuned Llama Model Access via API

By

–

27 July 2023 2h39

We are especially looking forward to providing access to fine-tuned versions of Llama as the community creates those. If you are experimenting with fine tuning Llama or any other model, it’s easy to use Poe as a frontend for your model via our API: https://
developer.poe.com/api-bots/quick
-start
…

Poe Hosts Fine-Tuned Llama Models for Developers

By

–

27 July 2023 2h39

We are also interested in helping host fine-tuned versions of Llama so that developers don’t have to manage a server or pay large costs if their bot gets popular. Please reach out to us at developers@poe.com if you are interested in participating in this program.

Meta Launches Llama 2 70b, 13b, 7b Models

By

–

27 July 2023 2h39

We are excited to add three new bots based on Llama 2: Llama-2-70b, Llama-2-13b, and Llama-2-7b. Our recommendation is to use 70b, as it provides the highest quality responses. It is significantly better than the 7b and 13b models that many people have been testing the last week.

Poe Platform Enables Llama Access on iOS Android

By

–

27 July 2023 2h39

We believe Poe is currently the only consumer product where you can use Llama with native iOS or Android apps, file uploading, or sharing and continuing conversations. You can find the bots at https://
poe.com/Llama-2-70b , https://
poe.com/Llama-2-13b , and https://
poe.com/Llama-2-7b.

→ View original post on X — @googleai

In-Context Learning Mechanisms in Transformers Explained

By

@googleai

–

27 July 2023 1h25

The mechanisms of in-context learning in transformers are not currently well understood. Visit the #ICML2023 Google Research booth today at 3:30 PM for a demonstration on how a type of gradient descent enables transformers to pay better attention to their context.

→ View original post on X — @hwchase17

More modular approach to running AI agents

By

@hwchase17

–

27 July 2023 1h05

this is very cool – more modular way to run agents

→ View original post on X — @aiatmeta

Meta AI Research Paper Wins ICML 2023 Outstanding Paper Award

By

@aiatmeta

–

27 July 2023 0h59

We’re honored to share that Learning-Rate-Free Learning by D-Adaptation, a paper by Meta AI research scientist @aaron_defazio & @konstmish was selected as an #ICML2023 Outstanding Paper! More details and a link to the paper