README.md

Training simple network on IRIS dataset

• Follow instruction to setup enviroment: DeepWorks samples

• Run train:

./samples/iris/run_iris_train.sh

Possible output:

Epoch: 0
Loss: 1.09987
Loss: 1.03461
Loss: 1.03461
Accuracy: 0.609375
Epoch: 1
Loss: 0.888799
Loss: 0.834636
Loss: 0.834636
Accuracy: 0.617188
Epoch: 2
Loss: 0.654118
Loss: 0.636887
Loss: 0.636887
Accuracy: 0.632812
Epoch: 3
Loss: 0.534014
Loss: 0.522762
Loss: 0.522762
Accuracy: 0.632812
Epoch: 4
Loss: 0.44576
Loss: 0.389611
Loss: 0.389611
Accuracy: 0.648438
Model saved: build/iris_model.bin

• Run test:

./samples/iris/run_iris_test.sh

Output:

Accuracy: 0.648438

You can also run samples using binary target directly:

• Run train:

./bin/sample_iris_train train <path-to-deepworks>/datasets/IRIS <batch_size> <num_epochs> <dump-frequency> <path-to-dump>

• Run test:

./bin/sample_iris_train test <path-to-deepworks>/datasets/IRIS <batch_size> <path-to-model>