Alex Tomala
@a__tomala
Followers
1K
Following
1K
Statuses
197
Research Engineer @GoogleDeepMind It’s time to think🫡
San Francisco, CA
Joined January 2023
@_sholtodouglas NotebookLM said its favorite equation from the book is 6ND because it’s a great rule of thumb to estimate runtimes :>
0
0
2
Jacob and the other authors on this book are among our absolute best performance optimizers at DeepMind. This is the best LLM optimization book for TPUs and it still has tons of useful insights for anyone using GPUs
Making LLMs run efficiently can feel scary, but scaling isn’t magic, it’s math! We wanted to demystify the “systems view” of LLMs and wrote a little textbook called “How To Scale Your Model” which we’re releasing today. 1/n
0
3
40
@evanjconrad @aidanogara_ AFAIK, your company seems to work better in situations where you don’t need constant compute and can cope with some interruptions, which is fine for training, but not inference.
0
0
2
@aidan_mclau May be interesting to try this benchmark on Chinese prompts with maybe a different embed model that specializes in Chinese
0
0
2
RT @nearcyan: Merry Christmas from Claude and I！ open digital presents from us at the domain in the image! 🎄
0
2
0
RT @nabla_theta: I decided to conduct an experiment at neurips this year: I randomly surveyed people walking around in the conference hall…
0
42
0
Excited to be at Neurips next week! Happy to chat about LLMs, DeepMind, and whatever cool topics you are into.
1
0
12
Big things are happening
Gemini
1
0
11
OpenAI trying to beat Google to an announcement just for Enrique to pull a fast one on them 🔥🔥🔥
JAX stocking stuffer: Symbolic JAX tracing based type-checking is working—sharding propagation coming soon! The time this will save…
0
0
14