Time-MoE

Visit Tool

Time-MoE is an open-source AI tool that provides billion-scale time series foundation models. It features a Mixture of Experts architecture and includes the Time-300B dataset for extensive time series analysis and forecasting.

Claim this tool

3Views

At a glance

Pricing

Open Source

Free tier

Yes

API

Skill level

Technical

About

What is Time-MoE?

Time-MoE is an open-source project offering a family of decoder-only time series foundation models, utilizing a Mixture of Experts architecture. These models are designed for auto-regressive operation, enabling universal forecasting with arbitrary prediction horizons and context lengths up to 4096. It scales up to 2.4 billion parameters and is trained from scratch. A key component is the Time-300B dataset, the largest open-access time series data collection, comprising over 300 billion time points across more than nine domains. Time-MoE supports making forecasts, fine-tuning with custom datasets in jsonl format, and evaluation on benchmark datasets, making it suitable for advanced time series analysis.

Best used for

Ideal for data scientists who need to develop and deploy large-scale time series foundation models, perform universal forecasting with long context lengths, and fine-tune models with custom datasets. Especially valuable for researchers and practitioners working with extensive time series data across multiple domains.

Common actions

forecast time series

analyze time series data

train foundation models

develop AI models

automated workflowcollaborationopen-sourceworkflowslow-code/no-codedeepfakeface swapping"AI Agents"github copilot

Capabilities

Key features

Billion-scale time series models
Mixture of Experts architecture
Time-300B dataset
Universal forecasting
Fine-tuning support
Auto-regressive operation

Target Audience

data scientist

Integrations

Not yet documented

Pricing & Plans

Open Source

Free

FAQs

What is the maximum sequence length Time-MoE supports for forecasting?

Time-MoE has a maximum sequence length (max_position_embeddings) of 4096 during training. For optimal forecasting performance, it is recommended that the sum of context_length and prediction_length does not exceed 4096. Longer sequence lengths can be supported by fine-tuning the model.

How can I prepare my dataset for fine-tuning Time-MoE?

To fine-tune Time-MoE, your dataset should be converted into a jsonl format. Each line must be a dictionary object with a 'sequence' field containing a list of time-series observations. You can save your data in jsonl, json, or pickle format.

Does Time-MoE include a pre-trained dataset?

Yes, Time-MoE includes the Time-300B dataset, which is the largest open-access time series data collection. It comprises over 300 billion time points across more than nine domains and is available on Hugging Face for direct use.

Trending

Subcategories trending in Coding & Development

Code Assistants DevOps & Infrastructure No-Code / Low-Code Testing & QA Backend & APIs Prompt Engineering

Trending

Also listed in

This tool also appears in

Data & Analytics › Predictive Analytics Data & Analytics › Statistical & Scientific

Explore

Browse AI tools by category

Content & Design Productivity & Business Coding & Development AI Agents & Automation Research & Education Wellness & Lifestyle Career Development Marketing & Growth Data & Analytics Customer Support & CX Finance E-commerce