Open Source

Engram

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models.

Source: GitHub Pricing: Open Source
💻 View Code

About This Project

This repository contains the official implementation for Engram, which explores conditional memory as a complementary sparsity axis to MoE. It modernizes classic N-gram embeddings for O(1) lookup to work directly alongside the Transformer backbone.

Tags

LLM memory moe sparsity

Installation & Setup

pip install torch numpy transformers sympy

Reviews & Ratings

Share your experience

User Reviews (0)

No reviews yet. Be the first to share your experience!