Llama3.1-Gutenberg-Doppel-70B Open-Source Large Language Model - Achieving Efficient Language Interaction through Fine-Tuning

Home

Llama3.1 Gutenberg Doppel 70B

Developed by nbeerbower

A large language model based on Hermes-3-Llama-3.1-70B-lorablated, fine-tuned on the Gutenberg dataset

Large Language Model

Transformers

#Literary Text Generation #High Instruction Following Accuracy #Multi-task Generalization

Downloads 424

Release Time : 10/11/2024

Model Overview

This model is a 70B-parameter large language model focused on text generation tasks, optimized on the Gutenberg dataset, suitable for various natural language processing scenarios.

Model Features

Gutenberg Dataset Fine-tuning

Specifically optimized on the jondurbin/gutenberg-dpo-v0.1 and nbeerbower/gutenberg2-dpo datasets

ORPO Tuning

Tuned for 3 epochs using the ORPO method, enhancing model performance

High Performance

Excellent performance across multiple evaluation tasks on the Open LLM Leaderboard

Model Capabilities

Text Generation

Instruction Following

Knowledge Q&A

Use Cases

Education

Reading Comprehension

Helps students understand complex text content

Research

Literature Analysis

Assists researchers in quickly analyzing large volumes of literature

🚀 Llama3.1-Gutenberg-Doppel-70B

This project is a fine - tuned model based on mlabonne/Hermes - 3 - Llama - 3.1 - 70B - lorablated, aiming to enhance text - generation capabilities through fine - tuning on specific datasets.

image/png

🚀 Quick Start

The Llama3.1-Gutenberg-Doppel-70B model is a fine - tuned version of mlabonne/Hermes - 3 - Llama - 3.1 - 70B - lorablated. It has been fine - tuned on jondurbin/gutenberg - dpo - v0.1 and nbeerbower/gutenberg2 - dpo.

✨ Features

Model Information

Property	Details
Library Name	transformers
Base Model	mlabonne/Hermes - 3 - Llama - 3.1 - 70B - lorablated
Datasets	jondurbin/gutenberg - dpo - v0.1, nbeerbower/gutenberg2 - dpo
License	llama3.1

Training Method

The model was ORPO tuned using 2x H100 for 3 epochs. Special thanks to Schneewolf Labs for providing the computing resources.

📚 Documentation

Open LLM Leaderboard Evaluation Results

The model's performance on the Open LLM Leaderboard is as follows:

Detailed results can be found here

Metric	Value
Avg.	35.68
IFEval (0 - Shot)	70.92
BBH (3 - Shot)	52.56
MATH Lvl 5 (4 - Shot)	13.75
GPQA (0 - shot)	12.64
MuSR (0 - shot)	22.68
MMLU - PRO (5 - shot)	41.52

Model Index Details

Model Name: Llama3.1 - Gutenberg - Doppel - 70B
Results:
- Task: Text Generation
  - Dataset: IFEval (0 - Shot)
    - Dataset Type: HuggingFaceH4/ifeval
    - Metrics: strict accuracy = 70.92
    - Source: Open LLM Leaderboard
  - Dataset: BBH (3 - Shot)
    - Dataset Type: BBH
    - Metrics: normalized accuracy = 52.56
    - Source: Open LLM Leaderboard
  - Dataset: MATH Lvl 5 (4 - Shot)
    - Dataset Type: hendrycks/competition_math
    - Metrics: exact match = 13.75
    - Source: Open LLM Leaderboard
  - Dataset: GPQA (0 - shot)
    - Dataset Type: Idavidrein/gpqa
    - Metrics: acc_norm = 12.64
    - Source: Open LLM Leaderboard
  - Dataset: MuSR (0 - shot)
    - Dataset Type: TAUR - Lab/MuSR
    - Metrics: acc_norm = 22.68
    - Source: Open LLM Leaderboard
  - Dataset: MMLU - PRO (5 - shot)
    - Dataset Type: TIGER - Lab/MMLU - Pro
    - Metrics: accuracy = 41.52
    - Source: Open LLM Leaderboard

📄 License

The model is under the llama3.1 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご