[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-9822":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":19,"compositeScore":20,"rankGlobal":10,"rankLanguage":10,"license":21,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":24,"hasPages":24,"topics":25,"createdAt":10,"pushedAt":10,"updatedAt":41,"readmeContent":42,"aiSummary":43,"trendingCount":16,"starSnapshotCount":16,"syncStatus":44,"lastSyncTime":45,"discoverSource":46},9822,"nixtla","Nixtla\u002Fnixtla","Nixtla","TimeGPT-1: production ready pre-trained Time Series Foundation Model  for forecasting and anomaly detection. Generative pretrained transformer for time series trained on over 100B data points. It's capable of accurately predicting various domains such as retail, electricity, finance, and IoT with just a few lines of code 🚀.","https:\u002F\u002Fwww.nixtla.io\u002Fdocs",null,"Jupyter Notebook",3908,325,35,48,0,11,12,33,29.54,"Other",false,"main",true,[26,27,28,29,30,31,32,33,34,35,36,37,38,39,40],"agent","agentic-ai","anomaly-detection","artificial-intelligence","deep-learning","forecasting","foundation-models","generative-ai-time-series","gpt","gpts","llm","llms","time-series","time-series-forecasting","timegpt","2026-06-12 02:02:13","# Nixtla   [![Tweet](https:\u002F\u002Fimg.shields.io\u002Ftwitter\u002Furl\u002Fhttp\u002Fshields.io.svg?style=social)](https:\u002F\u002Ftwitter.com\u002Fintent\u002Ftweet?text=Statistical%20Forecasting%20Algorithms%20by%20Nixtla%20&url=https:\u002F\u002Fgithub.com\u002FNixtla\u002Fneuralforecast&via=nixtlainc&hashtags=StatisticalModels,TimeSeries,Forecasting)  [![Slack](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FSlack-4A154B?&logo=slack&logoColor=white)](https:\u002F\u002Fjoin.slack.com\u002Ft\u002Fnixtlacommunity\u002Fshared_invite\u002Fzt-1pmhan9j5-F54XR20edHk0UtYAPcW4KQ)\n\n\u003Cdiv align=\"center\">\n\u003Cimg src=\"https:\u002F\u002Fraw.githubusercontent.com\u002FNixtla\u002Fneuralforecast\u002Fmain\u002Fnbs\u002Fimgs_indx\u002Flogo_new.png\"\u002F>\n\u003Ch1 align=\"center\">TimeGPT-1 \u003C\u002Fh1>\n\u003Ch3 align=\"center\">The first foundation model for forecasting and anomaly detection\u003C\u002Fh3>\n\n[![CI](https:\u002F\u002Fgithub.com\u002FNixtla\u002Fnixtla\u002Factions\u002Fworkflows\u002Fci.yaml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002FNixtla\u002Fnixtla\u002Factions\u002Fworkflows\u002Fci.yaml)\n[![PyPi](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fnixtla?color=blue)](https:\u002F\u002Fpypi.org\u002Fproject\u002Fnixtla\u002F)\n[![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache_2.0-blue.svg)](https:\u002F\u002Fgithub.com\u002FNixtla\u002Fnixtla\u002Fblob\u002Fmain\u002FLICENSE)\n[![docs](https:\u002F\u002Fimg.shields.io\u002Fwebsite-up-down-green-red\u002Fhttp\u002Fdocs.nixtla.io\u002F.svg?label=docs)](https:\u002F\u002Fdocs.nixtla.io)\n[![Downloads](https:\u002F\u002Fpepy.tech\u002Fbadge\u002Fnixtla)](https:\u002F\u002Fpepy.tech\u002Fproject\u002Fnixtla)\n[![Downloads](https:\u002F\u002Fpepy.tech\u002Fbadge\u002Fnixtla\u002Fmonth)](https:\u002F\u002Fpepy.tech\u002Fproject\u002Fnixtla)\n[![Downloads](https:\u002F\u002Fpepy.tech\u002Fbadge\u002Fnixtla\u002Fweek)](https:\u002F\u002Fpepy.tech\u002Fproject\u002Fnixtla)\n\n**TimeGPT** is a production ready, generative pretrained transformer for time series. It's capable of accurately predicting various domains such as retail, electricity, finance, and IoT with just a few lines of code 🚀.\n\n\u003C\u002Fdiv>\n\n## 🚀 Quick Start\n\nhttps:\u002F\u002Fgithub.com\u002FNixtla\u002Fnixtla\u002Fassets\u002F4086186\u002F163ad9e6-7a16-44e1-b2e9-dab8a0b7b6b6\n\n### Install nixtla's SDK\n\n```python\npip install nixtla>=0.7.0\n```\n\n### Import libraries and load data\n\n```python\nimport pandas as pd\nfrom nixtla import NixtlaClient\n```\n\n### Forecast using TimeGPT in 3 easy steps\n\n```python\n# Get your API Key at https:\u002F\u002Fnixtla.io\u002Ffree-trial?utm_source=nixtla.io&utm_campaign=\u002Fdocs\u002Freadme\n\n# 1. Instantiate the NixtlaClient\nnixtla_client = NixtlaClient(api_key = 'YOUR API KEY HERE')\n\n# 2. Read historic electricity demand data\ndf = pd.read_csv('https:\u002F\u002Fraw.githubusercontent.com\u002FNixtla\u002Ftransfer-learning-time-series\u002Fmain\u002Fdatasets\u002Felectricity-short.csv')\n\n# 3. Forecast the next 24 hours\nfcst_df = nixtla_client.forecast(df, h=24, level=[80, 90])\n\n# 4. Plot your results (optional)\nnixtla_client.plot(df, fcst_df, level=[80, 90])\n\n```\n\n![Forecast Results](.\u002Fnbs\u002Fimg\u002Fforecast_readme.png)\n\n### Anomaly detection using TimeGPT in 3 easy steps\n\n```python\n# Get your API Key at https:\u002F\u002Fnixtla.io\u002Ffree-trial?utm_source=nixtla.io&utm_campaign=\u002Fdocs\u002Freadme\n\n# 1. Instantiate the NixtlaClient\nnixtla_client = NixtlaClient(api_key = 'YOUR API KEY HERE')\n\n# 2. Read Data # Wikipedia visits of NFL Star (\ndf = pd.read_csv('https:\u002F\u002Fdatasets-nixtla.s3.amazonaws.com\u002Fpeyton-manning.csv')\n\n\n# 3. Detect Anomalies\nanomalies_df = nixtla_client.detect_anomalies(df, time_col='timestamp', target_col='value', freq='D')\n\n# 4. Plot your results (optional)\nnixtla_client.plot(df, anomalies_df,time_col='timestamp', target_col='value')\n```\n\n![AnomalyDetection](nbs\u002Fimg\u002Fanomaly.png)\n\n## 🤓 API support for other languages\n\nExplore our [API Reference](https:\u002F\u002Fdocs.nixtla.io) to discover how to leverage TimeGPT across various programming languages including JavaScript, Go, and more.\n\n## ❄️ Snowflake Deployment\n\nRun TimeGPT directly within your Snowflake environment. The deployment script creates stored procedures and UDTFs that enable forecasting and anomaly detection on your Snowflake data without moving it outside your infrastructure.\n\n```bash\npip install nixtla[snowflake]\npython -m nixtla.scripts.snowflake_install_nixtla\n```\n\nThe script will guide you through setting up external access integrations, configuring your API key, and deploying the forecasting components to your specified database and schema.\n\n## 🔥 Features and Capabilities\n\n- **Zero-shot Inference**: TimeGPT can generate forecasts and detect anomalies straight out of the box, requiring no prior training data. This allows for immediate deployment and quick insights from any time series data.\n\n- **Fine-tuning**: Enhance TimeGPT's capabilities by fine-tuning the model on your specific datasets, enabling the model to adapt to the nuances of your unique time series data and improving performance on tailored tasks.\n\n- **API Access**: Integrate TimeGPT seamlessly into your applications via our robust API. Upcoming support for Azure Studio will provide even more flexible integration options. Alternatively, deploy TimeGPT on your own infrastructure to maintain full control over your data and workflows.\n\n- **Add Exogenous Variables**: Incorporate additional variables that might influence your predictions to enhance forecast accuracy. (E.g. Special Dates, events or prices)\n\n- **Multiple Series Forecasting**: Simultaneously forecast multiple time series data, optimizing workflows and resources.\n\n- **Custom Loss Function**: Tailor the fine-tuning process with a custom loss function to meet specific performance metrics.\n\n- **Cross Validation**: Implement out of the box cross-validation techniques to ensure model robustness and generalizability.\n\n- **Prediction Intervals**: Provide intervals in your predictions to quantify uncertainty effectively.\n\n- **Irregular Timestamps**: Handle data with irregular timestamps, accommodating non-uniform interval series without preprocessing.\n\n## 📚 Documentation with examples and use cases\n\nDive into our [comprehensive documentation](https:\u002F\u002Fdocs.nixtla.io\u002Fdocs\u002Fgetting-started-timegpt_quickstart) to discover examples and practical use cases for TimeGPT. Our documentation covers a wide range of topics, including:\n\n- **Getting Started**: Begin with our user-friendly [Quickstart Guide](https:\u002F\u002Fdocs.nixtla.io\u002Fdocs\u002Fgetting-started-timegpt_quickstart) and learn how to [set up your API key](https:\u002F\u002Fdocs.nixtla.io\u002Fdocs\u002Fgetting-started-setting_up_your_api_key) effortlessly.\n\n- **Advanced Techniques**: Master advanced forecasting methods and learn how to enhance model accuracy with our tutorials on [anomaly detection](https:\u002F\u002Fdocs.nixtla.io\u002Fdocs\u002Ftutorials-anomaly_detection), fine-tuning models using specific loss functions, and scaling computations across distributed frameworks such as [Spark, Dask, and Ray](https:\u002F\u002Fdocs.nixtla.io\u002Fdocs\u002Ftutorials-computing_at_scale).\n\n- **Specialized Topics**: Explore specialized topics like [handling exogenous variables](https:\u002F\u002Fdocs.nixtla.io\u002Fdocs\u002Ftutorials-holidays_and_special_dates), model validation through [cross-validation](https:\u002F\u002Fdocs.nixtla.io\u002Fdocs\u002Ftutorials-cross_validation), and strategies for [forecasting under uncertainty](https:\u002F\u002Fdocs.nixtla.io\u002Fdocs\u002Ftutorials-uncertainty_quantification).\n\n- **Real World Applications**: Uncover how TimeGPT is applied in real-world scenarios through case studies on [forecasting web traffic](https:\u002F\u002Fdocs.nixtla.io\u002Fdocs\u002Fuse-cases-forecasting_web_traffic) and [predicting Bitcoin prices](https:\u002F\u002Fdocs.nixtla.io\u002Fdocs\u002Fuse-cases\u002Fbitcoin_price_prediction).\n\n## 🗞️ TimeGPT1 Revolutionizing Forecasting and Anomaly Detection\n\nTime series data is pivotal across various sectors, including finance, healthcare, meteorology, and social sciences. Whether it's monitoring ocean tides or tracking the Dow Jones's daily closing values, time series data is crucial for forecasting and decision-making.\n\nTraditional analysis methods such as ARIMA, ETS, MSTL, Theta, CES, machine learning models like XGBoost and LightGBM, and deep learning approaches have been standard tools for analysts. However, TimeGPT introduces a paradigm shift with its standout performance, efficiency, and simplicity. Thanks to its zero-shot inference capability, TimeGPT streamlines the analytical process, making it accessible even to users with minimal coding experience.\n\nTimeGPT is user-friendly and low-code, enabling users to upload their time series data and either generate forecasts or detect anomalies with just a single line of code. As the only foundation model for time series analysis out of the box, TimeGPT can be integrated via our public APIs, through Azure Studio (coming soon), or deployed on your own infrastructure.\n\n## ⚙️ TimeGPT's Architecture\n\nSelf-attention, the revolutionary concept introduced by the paper “Attention is all you need“, is the basis of the this foundational model. The TimeGPT model is not based on any existing large language model(LLMs). It is independently trained on vast timeseries dataset as a large transformer model and is designed so as to minimize the forecasting error.\n\nThe architecture consists of an encoder-decoder structure with\nmultiple layers, each with residual connections and layer normalization. Finally, a linear layer maps the decoder’s output to the forecasting window dimension. The general intuition is that attentionbased mechanisms are able to capture the diversity of past events and correctly extrapolate potential\nfuture distributions.\n\n![Arquitecture](nbs\u002Fimg\u002Fforecast.png)\n\nTimeGPT was trained on, to our knowledge, the largest collection of publicly available time series,\ncollectively encompassing over 100 billion data points. This training set incorporates time series\nfrom a broad array of domains, including finance, economics, demographics, healthcare, weather,\nIoT sensor data, energy, web traffic, sales, transport, and banking. Due to this diverse set of domains,\nthe training dataset contains time series with a wide range of characteristics\n\n---\n\n## ⚡️ Zero-shot Results\n\n### Accuracy\n\nTimeGPT has been tested for its zero-shot inference capabilities on more than 300K unique series, which involve using the model without additional fine-tuning on the test dataset. TimeGPT outperforms a comprehensive range of well-established statistical and cutting-edge deep learning models, consistently ranking among the top three performers across various frequencies.\n\n### Ease of use\n\nTimeGPT also excels by offering simple and rapid predictions using a pre-trained model. This stands in stark contrast to other models that typically require an extensive training and prediction pipeline.\n\n![Results](nbs\u002Fimg\u002Fresults.jpg)\n\n### Efficiency and Speed\n\nFor zero-shot inference, our internal tests recorded an average GPU inference speed of 0.6 milliseconds per series for TimeGPT, which nearly mirrors that of the simple Seasonal Naive.\n\n## 📝 How to cite?\n\nIf you find TimeGPT useful for your research, please consider citing the associated [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2310.03589):\n\n```\n@misc{garza2023timegpt1,\n      title={TimeGPT-1},\n      author={Azul Garza and Max Mergenthaler-Canseco},\n      year={2023},\n      eprint={2310.03589},\n      archivePrefix={arXiv},\n      primaryClass={cs.LG}\n}\n```\n\n## 🎉 Features and Mentions\n\nTimeGPT has been featured in many publications and has been recognized for its innovative approach to time series forecasting. Here are some of the features and mentions:\n\n- [TimeGPT Revolutionizing Time Series Forecasting](https:\u002F\u002Fwww.analyticsvidhya.com\u002Fblog\u002F2024\u002F02\u002Ftimegpt-revolutionizing-time-series-forecasting\u002F)\n- [TimeGPT: The First Foundation Model for Time Series Forecasting](https:\u002F\u002Ftowardsdatascience.com\u002Ftimegpt-the-first-foundation-model-for-time-series-forecasting-bf0a75e63b3a)\n- [TimeGPT: Revolutionising Time Series Forecasting with Generative Models](https:\u002F\u002Fmedium.com\u002F@22meera99\u002Ftimegpt-revolutionising-time-series-forecasting-with-generative-models-86be6c09fa51)\n- [TimeGPT on Turing Post](https:\u002F\u002Fwww.turingpost.com\u002Fp\u002Ftimegpt)\n- [TimeGPT Presentation at AWS Events](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=5pYkT0rTCfE&ab_channel=AWSEvents)\n- [TimeGPT: Machine Learning for Time Series Made Accessible - Podcast](https:\u002F\u002Fpodcasts.apple.com\u002Fbg\u002Fpodcast\u002Ftimegpt-machine-learning-for-time-series-made-accessible\u002Fid1487704458?i=1000638551991)\n- [TimeGPT on The Data Exchange](https:\u002F\u002Fthedataexchange.media\u002Ftimegpt\u002F)\n- [How TimeGPT Transforms Predictive Analytics with AI](https:\u002F\u002Fhackernoon.com\u002Fhow-timegpt-transforms-predictive-analytics-with-ai)\n- [TimeGPT: The First Foundation Model - AI Horizon Forecast](https:\u002F\u002Faihorizonforecast.substack.com\u002Fp\u002Ftimegpt-the-first-foundation-model)\n\n## 🔖 License\n\nTimeGPT is closed source. However, this SDK is open source and available under the Apache 2.0 License. Feel free to contribute (check out the [Contributing](https:\u002F\u002Fgithub.com\u002FNixtla\u002Fnixtla\u002Fblob\u002Fmain\u002FCONTRIBUTING.md) guide for more details).\n\n## 🏷️ Attribution\n\nNixtlaClient may be used to access services powered by technology from Google, Amazon, IBM, Datadog, and NXAI. All trademarks are the property of their respective owners.\n\n## 📞 Get in touch\n\nFor any questions or feedback, please feel free to reach out to us at ops [at] nixtla.io.\n","Nixtla\u002Fnixtla 是一个预训练的时间序列基础模型，用于预测和异常检测。该项目基于生成式预训练变压器架构，经过超过1000亿个数据点的训练，能够准确地应用于零售、电力、金融和物联网等多个领域。其核心功能包括时间序列预测和异常检测，并且只需几行代码即可实现复杂任务。TimeGPT-1 适合需要高效、准确时间序列分析的场景，如销售预测、能耗管理、金融市场分析以及设备监控等。",2,"2026-06-11 03:24:54","top_topic"]