The Data Weekly — 13.03.2023

Roman Kyslyi, PhD
The Data Weekly
Published in
Sent as a

Newsletter

1 min readMar 13, 2023
Photo by billow926 on Unsplash

Pandas 2.0 is here!
Biggest changes:
- Faster and more memory-efficient operations (added support for pyarrow in the backend)
- Copy-on-Write Optimisation (now it is similar to PySpark and how it executes code)
- Index can now hold NumPy numeric types (now it can be any NumPy numeric dtype, including int8, int16, int32, int64, uint8, uint16, uint32, uint64, float32, and float64.
Previously, only int64, uint64, and float64 types were supported.)

Alternative Chat LLMs are now available oh huggingface:
https://huggingface.co/spaces/olivierdehaene/chat-llm-streaming

PaLM-E a 562-billion parameter, general-purpose, embodied visual-language generalmodel is released: https://palm-e.github.io/

Prismer — a parameter-efficient vision-language model powered by an ensemble of domain experts is in opensource: https://twitter.com/DrJimFan/status/1633179734803890177
Seems that the next step would be large multimodal models.

Introducing Kaggle Models — after the Huggingface, Kaggle launches its own model hub.

Way to serve stable diffusion 3 times faster with Lightning — optimisations to serve model concurrently in production.

Also, we had a wonderful discussion about ChatGPT:

https://www.youtube.com/watch?v=-6g9tMwAGpY

That is all for the week, please subscribe and stay tuned!

--

--