What a 'Fused DeepLIFT' Actually Means: Gradient Overrides Without Hooks
jmschreiber91 · x · 2026-10-26
A technical explainer: DeepLIFT works by overriding the gradient operation to incorporate a reference value. General-purpose implementations attach hooks to model layers, leaving users unaware the model was modified. The fused version compiles the logic directly into the compute path (via tangermeme), eliminating hook overhead and yielding 7x speedups.
Related event: Fused DeepLIFT: Rewriting Attribution as a Computation Graph for 7x Speedup(4 posts)→
More from Research
- Open Models Match GPT-6 on MedMCQA but Trail 17 Points on Hard Clinical Cases — keunwoochoi · 2026-10-27
- Health-Focused Benchmark Reranks LLMs: Muse Spark Open Jumps from #9 to #1 — keunwoochoi · 2026-10-27
- Open-source Latch compiles PyTorch into training megakernels, 2.75x faster on H100 — sguada · 2026-10-27
- Goodfire commits $10M to a 'Human Genome Project for AI' to reverse-engineer frontier models — CatAstro_Piyush · 2026-10-27
- π0.5 study: raising demos from 10 to 438 lifts Place Ball in Bowl success from 7% to 83% — stingning · 2026-10-27
- HandsOn benchmark: 66 verified tasks and 18,000 human demos for dexterous manipulation — stingning · 2026-10-27