You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
All libraries are released on clojars. Cortex is not 1.0 yet preliminary and you should expect quite a few things to change
over time but it should allow you to train some initial classifiers or regressions. Note that the save format has not stabilized and although we do
just save edn data in nippy format it may require some effort to bring versions of saved forward.
Please see the various unit tests and examples for training a model. Specifically see:
mnist verification
Also, for an example of using cortex in a more real-world scenario please see:
mnist example.
Existing Framework Comparisons
Stanford CS 231 Lecture 12 contains a detailed
breakdown of Caffe, Torch, Theano, and TensorFlow.
TODO:
hdf5 import of major keras models (vgg-net). This requires each model along with a single input and per-layer outputs for that input. Please don't ask for anything to be supported unless you can provide the appropriate thorough test.
Recurrence in all forms. There is some work towards that direction in the compute branch and it is specifically designed to match the cudnn API for recurrence. This is less important at this point than running some of the larger pre-trained models.
Speaking of larger nets, multiple GPU support and multiple machine support (which could be helped by the above graph based description layer).
Profiling GPU system to make sure we are using as much GPU as possible in the single-gpu case.
Better data import/visualization support. We have geom and we have a clear definition of the datasets, now we need to put together the pieces and build some great visualizations as examples.
Getting Started:
Get the project and run lein test in both cortex and compute. The various unit tests train various models.
GPU Compute Install Instructions
Ubuntu
$ sudo apt install nvidia-cuda-toolkit
reboot
Install cuDNN and copy the cuDNN files to the corresponding folders in the local cuda installation (probably at /usr/local/cuda). For reference, follow the "Installing cuDNN" section here.
To check everything is working, run $ nvidia-smi
You should now have cuda8.0 installed. Current master is 8.0, so if you're running 7.5 you will need to change the javacpp dependency in your project file of the mnist Example.
Mac OS
These instructions follow the gpu setup from Tensor Flow, i.e.:
Should you see a jni linking error similar to this
Retrieving org/bytedeco/javacpp-presets/cuda/8.0-1.2/cuda-8.0-1.2-macosx-x86_64.jar from central
Exception in thread "main" java.lang.UnsatisfiedLinkError: no jnicudnn in java.library.path, compiling:(think/compute/nn/cuda_backend.c
lj:82:28)
at clojure.lang.Compiler.analyze(Compiler.java:6688)
at clojure.lang.Compiler.analyze(Compiler.java:6625)
at clojure.lang.Compiler$HostExpr$Parser.parse(Compiler.java:1009)
Make sure you have installed the appropriate CUDNN for your version of CUDA.