Davide

AI at Meta MSL

[Davide @DavideTortoise]

@AMD Same price and same memory bandwidth as the DGX spark. Why pay the AMD tax? You offer nothing compelling

[Davide @DavideTortoise]

@antirez The challenge is that Qwen wants to think for a very long time, so it often feels slower than it really is

[Davide @DavideTortoise]

@antirez @antirez could we not use product quantization on the Spark to shift inference towards being more compute-bound than memory-bound?

[Catherine Yeo @catherinehyeo]

Love seeing Naomi Osaka honor the CLRS Algorithms textbook at this year's Met Gala

[PyTorch @PyTorch]

Today Meta announced torchforge, a brand-new PyTorch-native library that makes it easy to use reinforcement learning (RL) to train AI agents.

Forge provides high-performance building blocks and ready-to-use examples, so you can focus on what’s novel about your use case rather than reinventing infrastructure. Built by the PyTorch team at Meta, Forge integrates seamlessly with environments, trainers, and deployment systems—bridging the gap between RL research and scalable, production-ready agentic development.

Read our blog to learn more: https://t.co/LEoF0ucwrf

#Forge #PyTorch #OpenSourceAI

[Davide @DavideTortoise]

@Vince2000_ @Martini373469 Can I have a code please?

[Davide @DavideTortoise]

Last year I got to work on post-training here at Meta. I noticed that Infra partners were fluent in pretraining, less in post-training. So I wrote a primer aimed at engineers: https://t.co/FOWU77xTka