8000 GitHub - djreiss/deepLearnPlayground: Playing around with deep learning
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

djreiss/deepLearnPlayground

Repository files navigation

My playground for dinking around with deep learning.

  • Notes:
  1. Keras documentation: http://keras.io/layers/core

To set up running using CPUs in parallel using intel optimized libs:

  1. install miniconda
  2. setenv CONDA_ENVS_PATH $MINICONDA/envs
  3. conda config --add channels intel
  4. conda create -n intelpython3 intelpython3_full python=3
  5. source activate intelpython3
  6. conda install keras tensorflow mkl-service

Or to use GPU, replace 2-6 above with:

  1. conda install matplotlib pillow scikit-learn
  2. conda install -c anaconda tensorflow-gpu=1.1.0
  3. conda install -c anaconda keras-gpu=2.0.2

Run times for notebook #1 (MNIST handwritten numbers) on GPU vs. multi-CPU:

  • GPU:
real	2m41.669s
user	2m54.096s
sys	0m12.324s
  • CPU (16 cores):
real	11m14.149s
user	149m24.364s
sys	5m40.696s

About

Playing around with deep learning

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published
0