# thuml/Large-Time-Series-Model

Official code, datasets and checkpoints for "Timer: Generative Pre-trained Transformers Are Large Time Series Models" (ICML 2024) and subsequent works

Repository: https://github.com/thuml/Large-Time-Series-Model
Canonical: https://ross.abutalabs.com/products/large-time-series-model
Homepage: https://arxiv.org/abs/2402.02368
Language: Python
License: MIT
License Family: permissive
Topics: large-models, time-series-analysis, transformer, time-series, foundation-models
Last push: 2026-03-22T04:11:35+00:00

## Health v2 (maintenance only)
Score: 58/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 73, release rhythm 35, longevity 65
- inputs: {"age_days": 917, "days_push": 164, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1010, forks 110 (observed 2026-08-28T04:03:12.959477+00:00)

## What it is
Official code, datasets, and checkpoints for Timer and Sundial, generative pre-trained Transformer foundation models for general time series analysis (ICML 2024, ICML 2025). It supports zero-shot forecasting, imputation, and anomaly detection via HuggingFace-compatible pretrained checkpoints.

## Use cases
- forecast time series with zero-shot pretrained models
- run generative probabilistic time series forecasting
- fine-tune a large time series model on my own data
- detect anomalies in time series data
- impute missing values in time series
- pre-train a time series foundation model
- download time series pre-training datasets

## When to choose
- you need zero-shot or few-shot forecasting without training a model from scratch
- you want a unified model for forecasting, imputation, and anomaly detection
- you are researching large time series foundation models
- your data is scarce and scenario-specific small models underperform

## When to avoid
- you need classical statistical forecasting like ARIMA or Prophet for simple univariate series
- you need a lightweight production model with minimal dependencies
- your task is not time series related

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, transformers, data-science
- domain: time-series, machine-learning, artificial-intelligence, data-science
- platform: python
- tags: time-series-forecasting, foundation-model, pretrained-model, zero-shot-forecasting, anomaly-detection, imputation, transformer, research-code, huggingface, gpu

## Member repositories
- thuml/Large-Time-Series-Model (main) score 58

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:12.959477+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:11:58.926599+00:00, confidence not recorded.
  - readme: https://github.com/thuml/Large-Time-Series-Model (fetched 2026-08-28T04:03:12.959477+00:00, sha d494b69fc10b)
  - homepage: https://arxiv.org/abs/2402.02368 (fetched 2026-08-29T13:12:06.832095+00:00, sha be6398e0382c)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T13:12:06.834987+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T13:12:06.838738+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T13:12:06.840663+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T13:12:06.836684+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.