Benchmark: figwig genomic data format loads GPU batches ~10x faster than bigWigs
jmschreiber91 · x · 2026-10-28
The core post of the thread: the new file format can do anything a bigwig can but is built for modern genomic ML, enabling GPU loading via a general-purpose kernel compatible with JAX, TensorFlow and PyTorch, plus fiber-seq support. The DataLoader benchmark table (64 windows x 2,048 bp and full reads) shows figwig with pinned memory or cuda+dlpack paths handling 512 tracks/2,048 windows in 1.3-1.7s versus 43s for bigWigs via pyBigWig — roughly an order of magnitude faster.
More from Infra
- 96.8% of eval labels came from the model itself — fixing bias cut costs 86% — AWS ML Blog · 2026-10-28
- Exowatt's fine-tuned Qwen3 models beat frontier models at up to 49× lower cost — AWS ML Blog · 2026-10-28
- NVIDIA Launches SimReady Central: 3,500+ Free OpenUSD Simulation Assets on Hugging Face — mishig25 · 2026-10-28
- eMudhra CA bust and PHP2Shell show shipping fast isn't operating well — rseroter · 2026-10-28
- Cadence Q3: recurring revenue up 18% as agents pull work through EDA tools — BenBajarin · 2026-10-28
- Cerebras partners with Oracle to bring its fast AI inference to OCI — vitaliychiley · 2026-10-28