DLStudio 1.0.4

An early version of an educational module that is being developed to make it easier to experiment with different deep learning networks in PyTorch

Consult the module API page at

https://engineering.purdue.edu/kak/distDLS/DLStudio-1.0.4.html

for all information related to this module, including information related to the latest changes to the code.

convo_layers_config = "1x[128,3,3,1]-MaxPool(2) 1x[16,5,5,1]-MaxPool(2)"
fc_layers_config = [-1,1024,10]

dls = DLStudio(
                  dataroot = "/home/kak/ImageDatasets/CIFAR-10/",
                  image_size = [32,32],
                  convo_layers_config = convo_layers_config,
                  fc_layers_config = fc_layers_config,
                  path_saved_model = "./saved_model",
                  momentum = 0.9,
                  learning_rate = 1e-3,
                  epochs = 2,
                  batch_size = 4,
                  classes = ('plane','car','bird','cat','deer','dog','frog','horse','ship','truck'),
                  use_gpu = True,
                  debug_train = 0,
                  debug_test = 1
              )

configs_for_all_convo_layers = dls.parse_config_string_for_convo_layers()
convo_layers = dls.build_convo_layers2( configs_for_all_convo_layers )
fc_layers = dls.build_fc_layers()
model = dls.Net(convo_layers, fc_layers)
dls.show_network_summary(model)
dls.load_cifar_10_dataset()
dls.run_code_for_training(model)
dls.run_code_for_testing(model)