OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Evaluation & Observability

Tfb

TFB is an open-source Python and PyTorch toolkit for benchmarking univariate and multivariate time-series forecasting methods using shared datasets, model baselines and experiment scripts.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Tfb — OpenViST project profile in AI Evaluation & Observability.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use TFB when you need a repository-based benchmark workflow for time-series forecasting and can prepare its dependencies and datasets locally.

Who it is for

  • Researchers comparing time-series forecasting methods.
  • Developers evaluating forecasting models within a shared benchmark workflow.

Documented uses

  • Train and evaluate forecasting baselines using repository-provided scripts.
  • Compare methods across prepared univariate and multivariate datasets.
  • Develop a forecasting method using the project's tutorial.
Before you choose

Requirements & limits

  • The README recommends Python 3.8 and identifies PyTorch 2.4.1.
  • The project offers both a Python dependency setup and a Dockerfile; Docker must be installed for the container route.
  • The README recommends requirements-docker.txt for reproducing scripts because it pins package versions.
  • Some retested algorithm results may differ from those in the paper.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up TFB and prepare its benchmark data

Install the project dependencies, choose the documented Docker setup if preferred, and place prepared datasets in the expected folder.

Before you start

  • A Python environment; the README recommends Python 3.8.
  • For the Docker route, an installed Docker service.

Steps

  1. For a Python installation, install dependencies with the command below; use requirements-docker.txt when aiming to reproduce the repository's experiment scripts with pinned package versions.

    Documented command
    pip install -r requirements.txt

  2. Alternatively, build the provided container with the command below and start it with the command below.

    Documented command
    docker build . -t tfb:latest
    Documented command
    docker run -it -v $(pwd)/:/app/ tfb:latest bash

  3. Download a preprocessed dataset from one of the sources listed in the README and place it under ./dataset.

Expected outcome

A prepared TFB environment with benchmark data in the location expected by its experiment scripts.

Guide limitations

  • The README recommends Python 3.8 and notes that version ranges in requirements.txt can affect reproduction.
  • The Docker route requires Docker; this setup guide does not run an experiment.
Documentation 1 ↗
Source-based practical guide

Run a documented multivariate forecasting benchmark script

Train and evaluate a model using a repository-provided multivariate forecasting experiment script.

Before you start

  • TFB dependencies installed; the README recommends Python 3.8.
  • A prepared dataset placed under ./dataset.
  • For closer reproduction of the repository scripts, the README recommends installing requirements-docker.txt.

Steps

  1. Use the setup and data-preparation instructions in the TFB README.

  2. Choose an experiment script from ./scripts/multivariate_forecast.

  3. Run the documented example with the command below.

    Documented command
    sh ./scripts/multivariate_forecast/ILI_script/DLinear.sh

Expected outcome

The selected repository experiment script runs the documented training and evaluation workflow for a multivariate forecasting model.

Guide limitations

  • The README warns that retested algorithm results may differ from those reported in the paper.
  • The guide does not establish results for other scripts, datasets, or environments.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
decisionintelligence/TFB · Official homepage title · retained in its source language
How we review sourcesSuggest a correction