A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.

Read and write spreadsheet files (CSV, XLSX and ODS), in a fast and scalable way

A fast, efficient universal vector embedding utility package.

A micro-library of stream components for building custom JSON and JSONC processing pipelines with a minimal memory footprint — parse, filter, and transform JSON far larger than available memory with a SAX-inspired token API, on Node.js or Web Streams.

Fastest and most memory efficient golang concurrent hashmap

An early-stage PHP parser designed for IDE usage scenarios.

Large Context Attention

A thread-safe queue faster and more resource efficient than golang's native channels

Professional backend framework for Node.js

Implementation of a memory efficient multi-head attention as proposed in the paper, "Self-attention Does Not Need O(n²) Memory"

APOLLO: SGD-like Memory, AdamW-level Performance; MLSys'25 Oustanding Paper Honorable Mention

VGGT-X: When VGGT Meets Dense Novel View Synthesis

Performance-oriented Data Structures for Kotlin

[CVPR 2025 Highlight] The official CLIP training codebase of Inf-CL: "Breaking the Memory Barrier: Near Infinite Batch Size Scaling for Contrastive Loss". A super memory-efficiency CLIP training scheme.

Easy Parallel Library (EPL) is a general and efficient deep learning framework for distributed model training.

Fast and memory-efficient immutable collections and helper data structures

Lightweight: Faster than light, Low on memory

Dragon: A Computation Graph Virtual Machine Based Deep Learning Framework.

Train Dense Passage Retriever (DPR) with a single GPU

Code for: "Social-Implicit: Rethinking Trajectory Prediction Evaluation and The Effectiveness of Implicit Maximum Likelihood Estimation" Accepted @ ECCV2022