# time-series-foundation-models/lag-llama

Lag-Llama: Towards Foundation Models for Probabilistic Time Series Forecasting

Repository: https://github.com/time-series-foundation-models/lag-llama
Canonical: https://ross.abutalabs.com/products/lag-llama
Language: Python
License: Apache-2.0
License Family: permissive
Topics: forecasting, foundation-models, time-series, time-series-forecasting, timeseries, timeseries-forecasting, llama, time-series-prediction, time-series-transformer, transformers, lag-llama
Last push: 2025-06-06T12:23:52+00:00

## Health v2 (maintenance only)
Score: 37/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 25, release rhythm 35, longevity 67
- inputs: {"age_days": 938, "days_push": 453, "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 1601, forks 203 (observed 2026-08-28T04:05:09.600872+00:00)

## What it is
Lag-Llama is the first open-source foundation model for probabilistic time series forecasting, built on a transformer architecture. It provides pretrained model weights for zero-shot forecasting as well as scripts for pretraining and finetuning.

## Use cases
- forecast future values of a time series zero-shot
- generate probabilistic forecasts with uncertainty estimates
- finetune a foundation model on my own time series data
- benchmark time series forecasting models
- replicate experiments from the Lag-Llama paper

## When to choose
- you need probabilistic forecasts without training a model from scratch
- you want a pretrained zero-shot time series model in Python
- you want to finetune or study a time series foundation model

## When to avoid
- you need classical statistical forecasting like ARIMA or Prophet
- you need a lightweight model for edge or low-resource deployment
- you need multivariate forecasting with rich covariates support

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, transformers
- domain: machine-learning, time-series, data-science
- platform: python
- tags: time-series-forecasting, foundation-model, probabilistic-forecasting, zero-shot-forecasting, lag-llama

## Member repositories
- time-series-foundation-models/lag-llama (main) score 37

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:09.600872+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 2026-08-30T03:52:49.519198+00:00, confidence not recorded.
  - readme: https://github.com/time-series-foundation-models/lag-llama (fetched 2026-08-28T04:05:09.600872+00:00, sha f3e36d64129e)
- Data as of 2026-08-30T08:39:29.467469+00:00.