ineshinspace

Open source libraries for MLX and Apple Silicon

mlxtaef

mlx-taef — Tiny AutoEncoders for diffusion latents in pure MLX. Decode FLUX.2 latents in ~260 ms or FLUX.1 in ~185 ms on M1 Max, vs ~2 s for the full VAE, with ~4× less peak memory.

More on PyPI GitHub

mlxteacache

mlx-teacache — TeaCache step-skipping for FLUX diffusion on Apple Silicon in pure MLX. Skip a calibrated subset of denoising steps; 1.44× on FLUX.1-dev at 25 steps.

More on PyPI GitHub

Denis Ineshin

Senior Full-stack Engineer at Booking.com

Amsterdam-based developer. Over a decade shipping open source — currently focused on MLX and Apple Silicon machine learning.

Buy me a coffee