Model Information Table

Name	Training Data	Params	Tokens	LR
TimeGEN-1	Time Series data from different domains	500M	100b	0.0001

Hardware and software

Training Factors
We used custom training libraries Nixtla’s open-source libraries, Nixtla’s Research Cluster and production clusters for pretraining. Fine-tuning and evaluation were also performed on third-party cloud compute.

Training Data

Overview
TimeGEN-1 was pretrained on 100 billion tokens of time series data from publicly available sources. Data Freshness
The pretraining data has a cutoff of September 2023 but some tuning data is more recent up to March 2024.

Risks and Limitations

• Accuracy: Artificial intelligence and machine learning are rapidly evolving fields of study. We are constantly working to improve TimeGEN-1 to make them more accurate, reliable, safe, and beneficial. Given the probabilistic nature of machine learning, the use of our Product may, in some situations, result in incorrect Output. You should always evaluate the accuracy of any Output as appropriate for your use case, including by using human review of the Output.
• Transferability between Domains: Time series models trained on data from different domains, may not always perform accurately when applied to a different domain.
• Impact of Extreme Events: Extreme events (such as natural disasters, economic crises, or pandemics) can significantly impact the accuracy of time series models. These events often create patterns and trends that the model has not encountered during training, leading to poor performance.
• Data Quality and Preprocessing: The quality of the input data greatly affects the accuracy of the time series model. Issues such as missing values, outliers, and inconsistent data can lead to unreliable predictions.

Mitigations: Ensure thorough data preprocessing, including cleaning, normalization, and imputation of missing values. Regularly audit the data pipeline to maintain high data quality.

How to use the model

Follow this article to deploy TimeGEN1 model with pay-as-you-go. Learn more about the Nixtla TimeGEN-1 model's request schema here .

Inference Samples

Sample Notebook	Description
Quick Start Forecast	Get started with forecasting using Nixtla’s TimeGEN1.
Finetuning	Fine-tuning is a powerful process for utilizing Time-GEN1 more effectively.
Anomaly Detection	Anomaly Detection involves monitoring ordered data points to spot irregularities that may signal issues or threats.
Exogenous Variables	Exogenous variables are external factors that can influence forecasts.
Demand Forecasting	Demand forecasting is the process of leveraging historical data and other analytical information to build models that help predict future estimates of customer demand for specific products over a specific period.

Sample Input

payload = {
    "freq": "D",
    "fh": 7,
    "y": {
        "2015-12-02": 8.71177264560569,
        "2015-12-03": 8.05610965954506,
        "2015-12-04": 8.08147504013705,
        "2015-12-05": 7.45876269238096,
        "2015-12-06": 8.01400499477946,
        "2015-12-07": 8.49678638163858,
        "2015-12-08": 7.98104975966596,
        "2015-12-09": 7.77779262633883,
        "2015-12-10": 8.2602342916073,
        "2015-12-11": 7.86633892304654,
        "2015-12-12": 7.31055015853442,
        "2015-12-13": 7.71824095195932,
        "2015-12-14": 8.31947369244219,
        "2015-12-15": 8.23668532271246,
        "2015-12-16": 7.80751004221619,
        "2015-12-17": 7.59186171488993,
        "2015-12-18": 7.52886925664225,
        "2015-12-19": 7.17165682276851,
        "2015-12-20": 7.89133075766189,
        "2015-12-21": 8.36007143564403,
        "2015-12-22": 8.11042723757502,
        "2015-12-23": 7.77527584648686,
        "2015-12-24": 7.34729970074316,
        "2015-12-25": 7.30182234213793,
        "2015-12-26": 7.12044437239249,
        "2015-12-27": 8.87877607170755,
        "2015-12-28": 9.25061821847475,
        "2015-12-29": 9.24792513230345,
        "2015-12-30": 8.39140318535794,
        "2015-12-31": 8.00469951054955,
        "2016-01-01": 7.58933582317062,
        "2016-01-02": 7.82524529143177,
        "2016-01-03": 8.24931374626064,
        "2016-01-04": 9.29514097366865,
        "2016-01-05": 8.56826646160024,
        "2016-01-06": 8.35255436947459,
        "2016-01-07": 8.29579811063615,
        "2016-01-08": 8.29029259122431,
        "2016-01-09": 7.78572089653462,
        "2016-01-10": 8.28172399041139,
        "2016-01-11": 8.4707303170059,
        "2016-01-12": 8.13505390861157,
        "2016-01-13": 8.06714903991011
    },
    "clean_ex_first": True,
    "finetune_steps": 0,
    "finetune_loss": "default"
}

Sample Output

{
  "timestamp": [
    "2016-01-14 00:00:00",
    "2016-01-15 00:00:00",
    "2016-01-16 00:00:00",
    "2016-01-17 00:00:00",
    "2016-01-18 00:00:00",
    "2016-01-19 00:00:00",
    "2016-01-20 00:00:00"
  ],
  "value": [
    8.05051040649414,
    7.909270763397217,
    7.8442702293396,
    8.21353816986084,
    8.600921630859375,
    8.293558120727539,
    8.150216102600098
  ],
  "input_tokens": 43,
  "output_tokens": 7,
  "finetune_tokens": 0
}