Hacker News: leschak

New comment by leschak in "Show HN: Autofit2 – End-to-end pipeline for multilingual text classification"

leschak — Sat, 27 Jun 2026 06:04:22 +0000

Yes, this is an alternative original implementation, from four years ago, when the concept of SetFit was still new and HF's project didn't exist. I guess its value nowadays lies in its simplicity. It is really simple. And practicable if you use TorchServe, because embeddings and classification model get serialized into one object.

New comment by leschak in "Migrating from Proxmox to NixOS and Incus"

leschak — Fri, 26 Jun 2026 07:39:31 +0000

It is a scripting language but for Mikrotiks you can scp a .rsc file and

  ssh $username@$routeraddr '/import file=setup.rsc'

Show HN: Autofit2 – End-to-end pipeline for multilingual text classification

leschak — Thu, 25 Jun 2026 13:58:21 +0000

Hi HN, Stefan here. autofit2 is a project I have been using at my previous company and is now opensourced. It has been used extensively in automated text moderation, but can be applied to any text/document classification task. We had success modeling offensive texts in 20+ languages (cf. github.com/neospe/dataload for all the datasets).

It's an integrated pipeline for lightweight multilingual text classification, covering preprocessing, training, and evaluation. It implements SetFit, a few-shot learning technique that works well for low-data regimes (down to a few dozen examples), and offers high throughput on CPUs, since it's based on Sentence Transformers. Dependencies are kept lean, but of course PyTorch itself isn't exactly small.

autofit2 takes a base model and a JSON config as input, and outputs a TorchServe model archive as well as a model card. The model card includes any benchmarks you have for your task, self-consistency tests, estimated CO2 emissions of the finetune, as well as an entropy-based bias analysis. For the bias eval, small test corpora for 50 languages are included. It works best with my EAR (Entropy-based Attention Regularization) fork of Sentence Transformers.

Feedback is welcome.

Comments URL: https://news.ycombinator.com/item?id=48673527

Points: 28

# Comments: 2

Autoops: Multi-region data and service mesh operated by a Makefile

leschak — Wed, 24 Jun 2026 11:45:17 +0000

Article URL: https://github.com/neospe/autoops

Comments URL: https://news.ycombinator.com/item?id=48658297

Points: 4

# Comments: 0

New comment by leschak in "I looked through attacks in my access logs"

leschak — Mon, 29 Jan 2024 16:14:27 +0000

cuts out like 80-90% of the automated scans in our cases.

New comment by leschak in "Ask HN: Can we do better than Git for version control?"

leschak — Sun, 10 Dec 2023 11:39:41 +0000

100% fossil. there has been a few threads.. and always someone points out edge cases that are only to be solved using git.. well i dont think so. you can actually go into the sqlite db and change stuff. i've recently started playing with its server api to direct user feedback from web to fossils ticketing system. it is just mature and feature packed and i honestly hope it will get as much recognition as sqlite someday.

New comment by leschak in "Gilles Deleuze – What is Philosophy? [audio]"

leschak — Sat, 09 Dec 2023 11:28:08 +0000

deleuze was actually pretty well read in the natural sciences of his times, especially physics. even if you find this short snippet vacuous-sounding, he is actually one of the few philosophers whose metaphors about impenetrable parts of reality are grounded.