Installation

Install NeuralForecast with pip or conda

You can install the released version of NeuralForecast from the Python package index with:

pip install neuralforecast

or

conda install -c conda-forge neuralforecast

Tip

Neural Forecasting methods profit from using GPU computation. Be sure to have Cuda installed.

Warning

We are constantly updating neuralforecast, so we suggest fixing the version to avoid issues. pip install neuralforecast=="1.0.0"

Tip

We recommend installing your libraries inside a python virtual or conda environment.

Extras

You can use the following extras to add optional functionality:

• distributed training with spark: pip install neuralforecast[spark]
• saving and loading from S3: pip install neuralforecast[aws]

Use our env (optional)

If you don’t have a Conda environment and need tools like Numba, Pandas, NumPy, Jupyter, Tune, and Nbdev you can use ours by following these steps:

1. Clone the NeuralForecast repo:

$ git clone https://github.com/Nixtla/neuralforecast.git && cd neuralforecast

1. Create the environment using astral’s uv:

$ pip install uv
$ uv venv --python 3.10

1. Activate the environment:

• On Linux/MacOS:

$ source .venv/bin/activate

• On Windows:

$ .\.venv\Scripts\activate

1. Install NeuralForecast Dev

$ uv pip install -e ".[dev]" --torch-backend cpu # for cpu backend
$ uv pip install -e ".[dev]" --torch-backend cu118 # for CUDA 11.8 PyTorch backend