Megabake compiler hits 5.5x torch.compile speedup on SmolLM
ricklamers · x · 2026-10-18
- Megabake is a PyTorch-native automated mega-kernel compiler, and author ayushsatyam146 shared new benchmark progress.
- Starting from gemma-2b at just 0.3x of torch.compile performance, the project now significantly beats it: 5.5x speedup on SmolLM 135M, which also served as the oracle model for experiments.
- Multiple other models tested — including Qwen variants and a DeepSeek drafter model — all gain meaningful speedups over torch.compile.
- ricklamers amplified the result with a "Just saying" nod, signaling it deserves community attention.
More from Infra
- Modal triples valuation in a week while data center owners raise nothing — YvesMulkers · 2026-10-18
- Reddit asks: is 2-bit or 3-bit quantization actually usable for daily LLM work? — Queasy_Back2588 · 2026-10-18
- Samsung and TSMC reportedly team up to produce Nvidia's next-gen HBM chips — Polymarket · 2026-10-18
- Safari Caps Incognito and Low-Power Mode at 30fps — Found via AI-Assisted Bench Bisect — oykun · 2026-10-18
- Running DeepSeek V4.1-Flash (318GB) on a single modded 4090 at 20+ tok/s — lxfater · 2026-10-18
- Hana Securities: YMTC likely to IPO this December, aiming to raise $33B — basedjensen · 2026-10-18