Today's digest · Friday, June 19

The 5 things in AI/dev today.

LiveNext issue at 7:00 CET
#1 / TODAY
Hugging Face·2 min·2d agoFREE

Beyond LoRA: Can you beat the most popular fine-tuning technique?

Hugging Face published an article exploring parameter-efficient fine-tuning (PEFT) techniques beyond LoRA. LoRA is noted as the most popular method for adapting large language models due to its efficiency. The post aims to evaluate alternative PEFT methods, including QLoRA, AdaLoRA, and DoRA, comparing their theoretical bases and practical implementations. This exploration seeks to expand the toolkit for AI developers, enabling more efficient model adaptation.

This exploration expands the toolkit for AI developers, enabling more efficient model adaptation with fewer resources.

peftlorafinetuningllms
huggingface.co
Beyond LoRA: Can you beat the most popular fine-tuning technique?
Putting a file in .gitignore does nothing if git already tracks it. I built a CLI to find the leftovers.
#2 / TOP STORY
DEV CommunityFREE

Putting a file in .gitignore does nothing if git already tracks it. I built a CLI to find the leftovers.

A new zero-dependency CLI tool, `gitslip`, has been developed to address a common Git issue where files already tracked by the repository are not ignored by subsequent `.gitignore` rules. This behavior can lead to sensitive data or large build artifacts remaining committed. `gitslip` identifies these "leftover" files that should be ignored but are still tracked, providing the necessary `git rm --cached` commands to untrack them, or applying the fix directly.

Is it agentic enough? Benchmarking open models on your own tooling
#3 / TOP STORY
Hugging FaceFREE

Is it agentic enough? Benchmarking open models on your own tooling

Hugging Face has published an article titled "Is it agentic enough? Benchmarking open models on your own tooling." The piece focuses on the process of evaluating the "agentic" capabilities of various open models. This benchmarking specifically involves testing these models against a user's or developer's custom-built tools. The aim is to determine how effectively open models can operate as agents within unique, application-specific environments, providing insights into their practical integration and performance.

aigest · daily

Get this every morning.

One email. The signal. Built for builders.

Free · Unsubscribe in one click · No trackers

// Worth knowing2 stories
// Yesterday18 stories