Distilling from weak labels: 1M images trained for $3.24
vanstriendaniel · x · 2026-08-18
The author presents a project extracting embellishments from the British Library book images dataset. The model was distilled entirely from weak labels with zero human annotations. The full 1M-image corpus training run cost only $3.24 in GPU compute. Links to the model, training data, and mask configuration are provided.
More from Infra
- Can RTX 5070ti (16GB) run Qwen 2.5 72B for coding agents? — zannix · 2026-08-18
- DeepSeek v4 local deployment hits 3000 t/s prefill on DGX Station — antirez · 2026-08-18
- Agent tracing becomes key for production debugging as Cloudflare adds support — krishnan · 2026-08-18
- CPO is 4.16x more energy efficient than pluggables, driving data center shift — casper_hansen_ · 2026-08-18
- Warp Factories Introduces Open Infrastructure for Cloud Software Factories — vikvang1 · 2026-08-18
- Nehemiah: On-demand Linux microVMs to hand to an AI — Rasmic · 2026-08-18