seriously shoutout to @dwhitena for consistently being on the side of low resource languages and the good work of @MasakhaneNLP (
https://
latent.space/p/practical-ai
-trends
…) – if any similar groups in latam/africa please make me aware of them
OPEN SOURCE
-
Shoutout to advocates for low resource languages and NLP
By
–
-
Predibase Releases New Features Including A100 Training and Mistral Support
By
–
Exciting News: we shipped a brand new #release last week that’s packed with awesome features and improvements: All training jobs run on #A100's by default Mistral-7b-instruct-v0.2 now live for inference + #finetuning New #deployments UI And more:
-
Document accessibility in open source JavaScript components
By
–
Honestly, if an open source JavaScript component doesn't include this level of detail on the accessibility story I tend to assume they haven't put the work in So if you've done that work, please document it like this!
-
FOSS Culture Toxic Behavior Creates Security Vulnerabilities
By
–
One striking thing about the xz backdoor is how the uncool, mean standards of behavior common in FOSS, that many of us decried for years, & that many defended as authentic, tough, etc., ended up being not just exclusionary loser behavior, but a significant attack surface.
-
PostgreSQL Development: Reliable Database Technology Evolution
By
–
Yes, and it's very cool technology and developing at a solid pace – it's still pretty new though, the PostgreSQL options are a lot more boring (in a good way)
-
Ruff Python Linter: Simplicity and Performance Benefits
By
–
I really like Ruff these days for both simplicity and performance, looks like it has a rule for this https://
docs.astral.sh/ruff/rules/mut
able-argument-default/
… -
Mini-Jamba: 69M Parameter Scaled-Down LLM Released
By
–
Someone made Mini-Jamba, a 69M parameter, scaled-down version of Jamba for testing. Very cool.
-
Using AI Magic Command in JupyterLab for Natural Language Chat
By
–
3. Use the "%%ai" magic command to specify a model chat with the model using a natural language prompt: Check this out: https://
github.com/jupyterlab/jup
yter-ai
… -
Low-Cost Privacy-First Project Reaches Technical Limits
By
–
I'm pretty much done with it I think – I think this is about as far as I can push it without adding a server-side component, and the whole point of this project is to avoid that and be as low-cost and privacy-friendly as possible
-
Reproducing Benchmarks: Torch Implementations and Performance Comparisons
By
–
The code to reproduce the benchmarks yourself is in the link. To note, the native torch implementations used here are well optimized implementations that are widely used. If you compare "plain" implementations (e.g. research code) it's not unusual to see a 5x speedup. I get the