ards Foundation Models for Probabilistic Time Series Forecasting 

 

 

 

Lag-Llama is the <b>first open-source foundation model for time series forecasting</b>! 

ined Time Series Model based on the Masked Encoder architecture. It is a universal time series forecasting model capable of addressing diverse forecasting tasks across multiple domains, frequencies, and variables in a zero-shot manner.  

 

This is a version of [Moirai small](https://huggingface.co/Salesforce/moirai-1.1-R-small) trained by Faculty AI. It was pre-trained on the [LOTSA data](https://huggingface.co/datasets/Salesforce/lotsa_data) using the [codebase](https://github.com/SalesforceAIResearch/uni2ts/tree/main/cli/conf/pretrain) provided by Woo et al. (2024). Both the dataset and codebase are licensed under the [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0). For more details on the model architecture, training, and results, please refer to the [paper](https://arxiv.org/abs/2402.02592). 

 

### Usage 

tion Time Series 

 

🚀 This is the official implementation of the NeurIPS 2024 paper "UniMTS: Unified Pre-training for Motion Time Series". 

 

<p align="center"> 

Tiny Time Mixer (TTM) Research-Use Model Card 

 

<p align="center" width="100%"> 

<img src="ttm_image.webp" width="600"> 

</p> 

This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration: 

- Library: [More Information Needed] 

- Docs: [More Information Needed] 

This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration: 

- Library: [More Information Needed] 

- Docs: [More Information Needed] 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

ined time series forecasting models** based on language model architectures. A time series is transformed into a sequence of tokens via scaling and quantization, and a language model is trained on these tokens using the cross-entropy loss. Once trained, probabilistic forecasts are obtained by sampling multiple future trajectories given the historical context. Chronos models have been trained on a large corpus of publicly available time series data, as well as synthetic data generated using Gaussian processes. 

 

For details on Chronos models, training data and procedures, and experimental results, please refer to the paper [Chronos: Learning the Language of Time Series](https://arxiv.org/abs/2403.07815). 

 

<p align="center"> 

TTM model, please see: https://huggingface.co/ibm-granite/granite-timeseries-ttm-v1 

 

ixer model, please see https://huggingface.co/ibm-granite/granite-timeseries-patchtsmixer 

 

M-R1 Model Card 

 

<p align="center" width="100%"> 

<img src="ttm_image.webp" width="600"> 

</p>