Cuda based library for fast and seamless deep learning inference.
Project description
Dolphin
General python package for CUDA accelerated deep learning inference.
- Documentation : ReadTheDoc
- Source code : https://github.com/MaximeDebarbat/Dolphin
- Bug reports : https://github.com/MaximeDebarbat/Dolphin/issues
- Getting Starterd :
It provides :
- A set of common image processing functions
- A TensorRT wrapper for easy inference
- Speeds up the inference with CUDA and TensorRT
- An easy to use API with Numpy
- A fast N-Dimensional array
Testing :
In order to test the package, you will need the library pytest
which you can run from the root of the project :
pytest
Install
pip install dolphin-python
Build
Dolphin can be installed with Pypi (coming soon) or built with Docker which is the recommended way to use it :
docker build -f Dockerfile \
--rm \
-t dolphin:latest \
.
Docker run
Ensure that you have the nvidia-docker
package installed and run the following command :
docker run \
-it \
--rm \
--gpus all \
-v "$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )":"/app" \
dolphin:latest \
bash
Please note that Dolphin might not work without the --gpus all
flag or --runtime nvidia
.
Acknowledgements
This project could not have been possible without PyCuda:
Andreas Klöckner, Nicolas Pinto, Yunsup Lee, Bryan Catanzaro, Paul Ivanov, Ahmed Fasih, PyCUDA and PyOpenCL: A scripting-based approach to GPU run-time code generation, > Parallel Computing, Volume 38, Issue 3, March 2012, Pages 157-174.
TODOs
- Improve
Engine
class in order to support int8 - Use Cython to speed up the code
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distributions
Built Distribution
Hashes for dolphin_python-0.0.10-py2.py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | f9586d73cb28824deecd4b25aeb0103d852b976d95737e5f5f9150b762511ca8 |
|
MD5 | b93fb0574f1f763dcbe417b9de3a1130 |
|
BLAKE2b-256 | 8b1442089d2689c6e92193e831ad94fd0d9907701943c0f026444a4e6392f874 |