Deep Water
What it is
- Native implementation of Deep Learning models for GPU-optimized backends (MXNet, Caffe, TensorFlow, etc.)
- State-of-the-art Deep Learning models trained from the H2O Platform
- Train user-defined or pre-defined Deep Learning models for image/text/H2OFrame classification from Flow, R, Python, Java, Scala or REST API
- Behaves just like any other H2O model (Flow, cross-validation, early stopping, hyper-parameter search, etc.)
- Deep Water is a legacy project (as of December 2017), which means that it is no longer under active development. The H2O.ai team has no current plans to add new features, however, contributions from the community (in the form of pull requests) are welcome.
Python/R Jupyter Notebooks
Check out a sample of cool Deep Learning Jupyter notebooks!
Pre-Release Downloads
This release of Deep Water is based on the latest H2O-3 release
The downloadable packages below are built for the following system specifications:
- Ubuntu 16.04 LTS
- NVIDIA Display driver at least 367
- CUDA 8.0.44 or later (we recommend the latest version) in /usr/local/cuda
- CUDNN 5.1 (placed inside of lib and include directories in /usr/local/cuda/)
To use the GPU, please set the following environment variables:
export CUDA_PATH=/usr/local/cuda
export LD_LIBRARY_PATH=$CUDA_PATH/lib64:$LD_LIBRARY_PATH
Python + Flow (most common)
R + Flow (R users)