🚀 We are pleased to announce that the Moulsot ASR model and dataset are now officially open source.
We invite the community to explore, experiment, and share valuable feedback to help us improve and evolve this work.
Your insights and contributions are highly appreciated.
Important Python memory tip you probably learned once and forgot:
x = x + y always creates a new object.
x += y tries to modify in place (if the type allows it).
Same syntax, very different memory behavior.
Matters a lot for performance with large data.
MoulSot-v0.3: the new Darija #ASR champ 👑
After weeks in the lab at #AtlasIA 🧪, results are in; matching and even slightly beating the ISMA model from the Ministry of Digital Transition & Administration Reform in some benchmarks.
#AtlasIA#MoulSot
Back to writing :)
5-post series moving through science, perception, memory, language, and the mind itself, asking how much of what we call knowing actually stands once you press on it.
Happy reading!
https://t.co/6VUZPcG5Bm
I created documentation over Claude Code's Codebase, which explains
- Its pipeline
- How it works
- How it handles Context
- How it handles Memory
& More
Read it here - https://t.co/GngrSvWAmh
Newton and Hooke's rivalry: In 1675, Newton wrote he had seen further "by standing on the shoulders of giants". A line that sounded humble but was a jab at Hooke, who was short and physically deformed, implying Newton relied on "giants", not men like him
Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: https://t.co/CDSQ8HpZoc
LiteLLM HAS BEEN COMPROMISED, DO NOT UPDATE. We just discovered that LiteLLM pypi release 1.82.8. It has been compromised, it contains litellm_init.pth with base64 encoded instructions to send all the credentials it can find to remote server + self-replicate. link below