Average Ratings 0 Ratings
Average Ratings 0 Ratings
Description
TabFM is an innovative zero-shot foundation model specifically created for handling tabular data, aimed at streamlining classification and regression processes that usually necessitate extensive manual model training, hyperparameter optimization, and tailored feature engineering. By transforming the challenge of tabular prediction into an in-context learning task, TabFM avoids the need to train a new supervised model for every dataset; instead, it consolidates historical training examples and target testing rows into a single cohesive prompt, allowing it to discern the intricate relationships between various columns and rows during inference. Given that tables are inherently two-dimensional and do not rely on a specific order, TabFM employs a hybrid architecture that integrates alternating attention mechanisms for both rows and columns, row compression techniques, and a specialized Transformer designed for in-context learning based on these compressed row embeddings. This sophisticated framework enables the model to effectively capture complex interactions and dependencies among features while maintaining computational efficiency, particularly advantageous for processing larger datasets. Furthermore, this approach not only enhances performance but also significantly reduces the time and resources typically required for model development in tabular data tasks.
Description
Timestral represents a suite of time-series foundation models along with a Python inference toolkit designed to create multivariate probabilistic forecasts based on historical data sequences. It comes in four different model sizes—5M, 28M, 123M, and 209M parameters—that are capable of handling past-only, known-future, and categorical covariates, providing quantile forecasts while accommodating context windows as extensive as 32,768 time steps. Users have the flexibility to load pretrained weights either from the Hugging Face Hub or directly from local files, enabling them to perform inference on both CPU and GPU. Such forecasts can be seamlessly integrated into various applications, including monitoring systems, capacity planning, energy management, supply chain logistics, financial risk assessment, or maintenance operations. Furthermore, Timestral offers zero-shot forecasting capabilities and can be tailored to specific domain data through the Mistral Forge. This package is primarily aimed at data scientists, machine learning engineers, and collaborative teams focused on developing robust forecasting solutions; the repository code is licensed under Apache 2.0, while the model weights are governed by a separate Modified MIT license. With its comprehensive features, Timestral is positioned to significantly enhance the accuracy and efficiency of predictive analytics in numerous industries.
API Access
Has API
No
API Access
Has API
Yes
Screenshots View All
No images available
Integrations
Mistral Forge
No
Pricing Details
Free
Free Trial
No
Free Version
Yes
Pricing Details
No price information available.
Free Trial
No
Free Version
Yes
Deployment
Web-Based
Yes
On-Premises
No
iPhone App
No
iPad App
No
Android App
No
Windows
No
Mac
No
Linux
No
Chromebook
No
Deployment
Web-Based
No
On-Premises
Yes
iPhone App
No
iPad App
No
Android App
No
Windows
No
Mac
No
Linux
No
Chromebook
No
Customer Support
Business Hours
No
Live Rep (24/7)
No
Online Support
Yes
Customer Support
Business Hours
No
Live Rep (24/7)
No
Online Support
Yes
Types of Training
Training Docs
Yes
Webinars
No
Live Training (Online)
No
In Person
No
Types of Training
Training Docs
Yes
Webinars
No
Live Training (Online)
No
In Person
No
Vendor Details
Company Name
Founded
1998
Country
United States
Website
research.google/blog/introducing-tabfm-a-zero-shot-foundation-model-for-tabular-data/
Vendor Details
Company Name
Mistral AI
Founded
2023
Country
France
Website
github.com/mistralai/timestral
Product Features
Product Features
Alternatives
Alternatives
No Alternatives