FastEstimator is a high-level deep learning API. With the help of FastEstimator, you can easily build a high-performance deep learning model and run it anywhere. 😉
- Python3
- TensorFlow 1.12.0
- Horovod (Only needed for distributed training)
pip install fastestimator
Docker container creates isolated virtual environment that shares resources with host machine. Docker provides an easy way to set up FastEstimator running environment, users can either build from dockerfile or pull from Docker Hub.
- GPU:
docker build -t fastestimator_gpu -f docker/Dockerfile.gpu .
- CPU:
docker build -t fastestimator_cpu -f docker/Dockerfile.cpu .
Pull Image from Docker Hub
- GPU:
docker pull fastestimator/fastestimator:latest-gpu
- macOS/Linux:
$ fastestimator train --entry_point tutorial/mnist.py
-
Windows, macOS and Linux:
check out this notebook
$ ./fastestimator/test/tests.sh
For more info about building models and training, check out FastEstimator documentation