Llama-68m Open-Source Small Language Model - Empowering Predictive Reasoning Research, Free to Use!

Llama 68m

Developed by JackFram

This is a small LLaMA-like language model with 68 million parameters, primarily used for speculative inference research in the SpecInfer paper.

Downloads 573.35k

Release Time : 7/19/2023

Model Overview

A lightweight language model trained on Wikipedia and partial C4 dataset, developed as the base speculative model for the SpecInfer paper.

Lightweight Design

Compact model with only 68 million parameters, ideal as a base model for speculative inference

Multi-source Training

Trained on combined datasets including Wikipedia, C4-en and C4-realnewslike

Research-oriented

Specifically designed for speculative inference and token tree verification research in the SpecInfer paper

English text generation

Academic Research

Speculative Inference Research

Serves as the base small speculative model in the SpecInfer paper

Used to validate the effectiveness of speculative inference and token tree verification methods

Lightweight Applications

Text Generation in Resource-constrained Environments

Provides basic text generation capability under limited computational resources

Property	Details
Model Type	LLaMA - like model
Training Data	Wikipedia, part of the C4 - en and C4 - realnewslike datasets
Pipeline Tag	text - generation

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base