Install CUDA -> https://developer.nvidia.com/cuda-downloads
Install cuDNN -> https://developer.nvidia.com/cudnn
https://www.anaconda.com/download/
Effsubsee's model runs on Python 2.7.
> conda create --name effsubsee python=2.7
> source activate effsubsee
> conda install pandas scikit-learn h5py tqdm tabulate
Effsubsee uses Pytorch version 0.1.12, which is not the latest version.
If you're running on Mac OSX, you need to install pytorch from source if you want GPU support.
If you only want to run this on the CPU, you can install with pip.
> pip install http://download.pytorch.org/whl/torch-0.1.12.post2-cp27-none-macosx_10_7_x86_64.whl
If you're running on Linux, you can pip install for both GPU and CPU options.
> pip install http://download.pytorch.org/whl/cu80/torch-0.1.12.post2-cp27-cp27mu-manylinux1_x86_64.whl
Install torchvision and ibmseti.
> pip install torchvision==0.1.8
> pip install ibmseti
Download the simulation data from here. Any of the data sets will work, but try the preview test set to start with.
> wget https://dal.objectstorage.open.softlayer.com/v1/AUTH_cdbef52bdf7a449c96936e1071f0a46b/simsignals_v3_zipped/primary_testset_preview_v3.zip
> git clone https://github.com/sgrvinod/ml4seti-Effsubsee
> cd ml4seti-Effsubsee
The following calls an example python script that utilizes one of the sets of model parameters that were calculated by the Effsubsee team.
This example runs on the CPU.
> python single_shot_cpu.py wresnet34x2\ models/wresnet34x2\ FOLD1/FOLD1_BEST_wresnet34x2_batchsize96_checkpoint.pth.tar folds/mean_stddev_primary_full_v3__384t__512f__logmod2-ph.hdf5 /path/to/primary_testset_preview_v3/00b3b8fdb14ce41f341dbe251f476093.dat
You can follow the instructions in the repository README to run the test.py script, which will run on your GPU.
Install CUDA -> https://developer.nvidia.com/cuda-downloads
Install cuDNN -> https://developer.nvidia.com/cudnn
https://www.anaconda.com/download/
Signet's code runs on Python 3.x.
> conda create --name signet python=3.5
> source activate signet
> conda install scikit-learn numpy tabulate
Signet's model uses the latest stable version of pytorch (0.2.0) and torchvision (0.1.9).
Install Pytorch without GPU support
> conda install pytorch torchvision -c soumith
OR install Pytorch with GPU support
> conda install pytorch torchvision cuda80 -c soumith
You'll need to install Pytorch from source if you want GPU support. Otherwise, to install the pytorch version 0.2.0 (as of September 2017)
> conda install pytorch torchvision -c soumith
> pip install ibmseti==2.0.0.dev5
Download the simulation data from here. Any of the data sets will work, but try the preview test set to start with.
> wget https://dal.objectstorage.open.softlayer.com/v1/AUTH_cdbef52bdf7a449c96936e1071f0a46b/simsignals_v3_zipped/primary_testset_preview_v3.zip
NB: I've cloned sagelywizard's repository to add some scripts. If my updates are added to his code base, this note will be removed.
> git clone https://github.com/gadamc/ml4seti ml4seti-signet
> cd ml4seti-signet
> wget https://dal.objectstorage.open.softlayer.com/v1/AUTH_cdbef52bdf7a449c96936e1071f0a46b/code_challenge_models/signet/final_densenet_model.pth
The following calls an example python script that will perform a signal classification using the CPU.
> python single_shot_cpu.py /path/to/primary_testset_preview_v3/00b3b8fdb14ce41f341dbe251f476093.dat final_densenet_model.pth
To use the GPU, you can follow the instructions in the README.
Hi!
is there any chance to update those links to the datasets?
Thanks in advance :)