← sources

Noise: Speedrun for pretraining without backpropagation - 4000x faster

source: reddit (reddit:StableDiffusion)published 2026-10-23 12:22ingested 2026-10-24 02:38

retained text

[qlabs-eng/noise: Speedrun for pretraining without backpropagation](https://github.com/qlabs-eng/noise)

[Noise - Q](https://qlabs.sh/noise#update)

  
[Noise](https://github.com/qlabs-eng/noise) is a pretraining speedrun for learning algorithms that never run a backward pass. Model, data and score are fixed; two tracks ask how low the loss can go and how fast. Our [Dust](https://qlabs.sh/research/dust/) baseline sits at 4.167.

https://preview.redd.it/j82721i8m7xh1.png?width=1200&format=png&auto=webp&s=23af795a548d4c1f62ecd07b2dab73080c5d0e15

In its first ten days, the community made Noise's speedrun over 4,000x faster: the time to reach validation loss 4.2 without a backward pass fell from 1.76 h to 1.6 s. We wrote that another 1,000x on top of Dust was possible. We didn't expect the speedrun to pass that in ten days, and we're grateful to everyone who submitted.

https://preview.redd.it/pzi7lmpbm7xh1.png?width=1200&format=png&auto=webp&s=f206ab33364bf8fe16ecff358735ff445b455a3c

  


document fetched 2026-10-24 10:34 · canonical https://reddit.com/r/StableDiffusion/comments/1xdciou/noise_speedrun_for_pretraining_without