Our revolutionary AI Model Develop Tool
Mod2 AI 

It is an innovative tool designed to simplify and optimize the development and deployment process of artificial intelligence models.

Request a Demo

Mod2 AI

A platform aimed at making AI technology more accessible, affordable, and secure by providing localized, efficient, and secure AI development tools to help users overcome challenges in the field of AI.

Start Now

Shortage and high cost of local GPUs

Increasing demand for AI capabilities


Data security issues in cloud computing

Cost-effective AI solutions

Customization and flexibility


 The bottlenecks encountered by AI development at the present stage

Service Goals

  • Make AI development more economical, efficient and safe.
  • Support small and medium-sized enterprises to easily adopt AI technology.  
closeup photo of turned on smartphone near keyboard

Software positioning

  • As a multifunctional AI development platform, it provides localized model training and deployment.
  • The interface is user-friendly and adaptable to various business needs.
sticky notes on paper document beside pens and box

Purpose

  • Promote the popularization and application of AI technology.
  • Help enterprises improve efficiency and enhance competitiveness.

Low Cost Local Training

Easy-to-use UI

Data Privacy​

Customizable ecosystem

Spec & Benchmark

Localized solution, processing data in a private environment to enhance data security.

Fine-tuning strategy: LoRA with FP16 | Dataset: Alpaca_data_en_52k

Compatible with Windows Subsystem for Linux (WSL) and Windows. Performance may vary.

236B
Model Fine Tuned
80%+
Cost Saving
∞
nodes for cluster computing
2 x 2
2-tiers Offloading * 2-tiers Cluster
Aug，2024
Mod2 1.0
Nov，2024
Mod2 2.0

Inference

Multi-Node Cluster

Multi-Datasets + LMM

Mod2 2.1

YOLO, ImageNet, CIFAR

DC-GAN, stable diffusion

K-FOLD

RL

Logistic Regression

Linear Regression

Future
Mod2 3.0

Tracking System Web/App:

Training status tracking

Experiment configs

Remote Operation
Window: Light version

Mod2

​236B

LLM Backbone

  • Offloading strategies：
    VRAM Only / VRAM + DRAM / VRAM + DRAM + SSD
  • GPU Type：NVIDIA / AMD
  • Fine-tuning type：Full / Freeze / Lora / QLora
  • Support GPU：NVIDIA / AMD
  • Time scheduling：
  • Resume_from_checkpoint：
  • Expert mode：

Others

70B

LLM Backbone

  • Offloading strategies：
    VRAM + DRAM + SSD (N-N only)
  • GPU Type：NVIDIA
  • Fine-tuning type：Full
  • Support GPU：NVIDIA
  • Time scheduling：
  • Resume_from_checkpoint：
  • Expert mode：
  • Progress bar：

Pre-Trained Models