# qlabs-eng/noise

> A speedrun benchmark for pretraining a transformer without backpropagation: fixed model, data and metric, with participants competing on forward-only learning algorithms.

- Magnitude: 2.0 out of 10 — Steady
- Stars: 115 total · +7 stars measured 2026-10-18, 15:52–18:00 UTC, ≈ 9 by evening
- Star trust: star growth looks organic
- Category: Language models · Language: Python · License: MIT · Created: 2026-10-12 · Last push: 2026-10-18
- GitHub: https://github.com/qlabs-eng/noise · Page: https://gitnova.dev/en/r/qlabs-eng/noise

## Useful for

- Test a forward-only learning algorithm on the fixed transformer and FineWeb dataset
- Compare time-to-4.2 validation loss against current speedrun records
- Reproduce the Dust and EGGROLL baselines on a single 8xH100 node

## Why it’s here

- Star-counter measurements on 2026-10-18 (UTC), 15:52–18:00: 108 → 115 stars (+7). This is the change over that interval.
- Estimated end-of-day forecast: about +9 stars, using observed gains and the previous day.
- The repository is 6 days old and already has 115 stars. With less than two weeks of history, there's no usual pace to compare the spike against yet.
- Top new repositories this week: #191.
- Recent forks include notable developers: @mvanhorn (5,529 followers).

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 11
- Issues and pull requests: 12
- Watchers: 3
- Average over the last week: 13 per day
- Usual pace: too little history (under two weeks)
- Stars in the last hour (measured): 4

## Stars per day, last 7 days (oldest → newest, today is partial)

2026-10-12 … 2026-10-18: 38, 14, 3, 9, 6, 11, 7

## Spotted in now

- Top new repositories this week: #191

## Similar by description

1. **qlabs-eng/dust** — 1.0 · Steady · Language models · Python · +2 stars measured 2026-10-18, 00:28–18:03 UTC, ≈ 3 by evening
   A minimal implementation of research on training a transformer with forward evaluations and SGD without backpropagation, plus a backprop baseline. For researchers exploring alternatives to backprop.
   Full card: https://gitnova.dev/en/r/qlabs-eng/dust.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-10-18 18:06 UTC, updated every 30 minutes.