GIGAZINE

A new inference engine called 'Colibrì' has emerged that can run the massive AI 'GLM-5.2,' with 744 billion parameters, on a regular PC with 25GB of memory.

Colibrì, an inference engine that runs the large-scale language model 'GLM-5.2,' which has 744 billion parameters, on a general-purpose PC with 25GB of memory, has been released. Instead of loading ...
note

Eliminating 'Over-engineering' in Python: Simplifying Implementations with functools.cache Instead of SQLite

This time, I would like to organize some thoughts on efficient cache implementation in Python and the trap of 'over-engineering' we often fall into, based on the article I Rewrote My Python Script ...
note

AION-CORE 64 | SSD Streaming a General Reasoner with 85.8GB Hy3 on a 64GB M1 Max

IO | This is your AI Navigator. Yesterday, on AION-CORE 64, I launched the 358GB-class GLM-5.2 using Colibri. It calls only the necessary parts from the SSD for a massive intelligence that does not ...
GitHub

Deep Reinforcement Learning-Based Cache Replacement Policy

We adopt a learning-based method to cache replacement strategy, aiming to improve the miss rate of existing traditional cache replacement policies. The main idea of modeling is to regard the strategy ...
GitHub

ARDY: Autoregressive Diffusion with Hybrid Representation for Interactive Human Motion Generation

ARDY is an autoregressive diffusion model designed for interactive motion generation, supporting online text prompting and flexible long-horizon kinematic constraints (root paths/waypoints, full-body ...