LLM-And-More  by IceBearAI

LLM trainer and app builder for complete workflow from data to service

created 1 year ago
384 stars

Top 75.6% on sourcepulse

GitHubView on GitHub
Project Summary

LLM-And-More provides a comprehensive, plug-and-play solution for training and deploying Large Language Models (LLMs). It targets professional developers and business users, aiming to simplify the entire LLM workflow from data preparation to service deployment, offering best practices and performance optimizations.

How It Works

The platform is structured into six modular components: Data, Training, Monitoring, Evaluation, Deployment, and Interaction. This modular design allows for a systematic approach to LLM application development. Key advantages include integrated data annotation with quality checks, a high-performance training framework leveraging DeepSpeed for distributed training, and real-time monitoring of resource usage and model performance metrics. The system aims to abstract away complex deep learning intricacies, providing intelligent defaults and actionable insights for users.

Quick Start & Requirements

  • Installation: Clone the repository, install Docker and NVIDIA Container Toolkit. For local development, ensure Go 1.22+ is installed and run go mod tidy.
  • Deployment: Use docker-compose up for full service deployment.
  • Prerequisites: Docker, NVIDIA Container Toolkit, Go (for local dev).
  • Resources: Requires GPU-enabled nodes for training and inference.
  • Documentation: Model Fine-tuning Guide

Highlighted Details

  • Supports multiple LLM architectures including LLaMA, LLaMA-2, Baichuan2, ChatGLM3, and Qwen, with options for full parameter or LoRA fine-tuning.
  • Offers specialized solutions for FAQ and General scenarios, with plans for RAG, Creative Writing, and Agent scenarios.
  • The FAQ scenario claims over 30% improvement in response accuracy by focusing on intent recognition.
  • Deployment is designed for flexibility across various Kubernetes environments (private, cloud).

Maintenance & Community

  • Community interaction is encouraged via a WeChat group.

Licensing & Compatibility

  • The repository does not explicitly state a license in the README.

Limitations & Caveats

  • Deployment and Interaction modules are marked as "Coming Soon."
  • RAG, Creative Writing, and Agent scenarios are also "Coming Soon."
  • The lack of a specified license may pose compatibility issues for commercial or closed-source use.
Health Check
Last commit

1 year ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
7 stars in the last 90 days

Explore Similar Projects

Starred by Chip Huyen Chip Huyen(Author of AI Engineering, Designing Machine Learning Systems), Pietro Schirano Pietro Schirano(Founder of MagicPath), and
1 more.

SillyTavern by SillyTavern

3.2%
17k
LLM frontend for power users
created 2 years ago
updated 3 days ago
Feedback? Help us improve.