def get_args(filename, sd, bn, im):
x = ArffToArgs()
x.set_input(filename)
x.set_class_index("last")
x.set_impute(im)
x.set_binarize(bn)
x.set_standardize(sd)
args = x.get_args()
x.close()
return args
def get_args(filename, sd, bn, im):
x = ArffToArgs()
x.set_input(filename)
x.set_class_index("last")
x.set_impute(im)
x.set_binarize(bn)
x.set_standardize(sd)
args = x.get_args()
x.close()
return args
bs = args["batch_size"]
else:
bs = 128
X_test = args["X_test"]
preds = iter_test(X_test).tolist()
new = []
for pred in preds:
new.append(np.eye(args["num_classes"])[pred].tolist())
return new
if __name__ == '__main__':
x = ArffToArgs()
x.set_input("data/cpu_act.arff")
x.set_class_index("last")
x.set_impute(True)
x.set_binarize(True)
x.set_standardize(True)
x.set_arguments(
"adaptive=True;alpha=0.01;lambda=0;epochs=500;rmsprop=True")
args = x.get_args()
#args["debug"] = True
args["X_test"] = np.asarray(args["X_train"], dtype="float32")
model = train(args)
test(args, model)
if "batch_size" in args:
bs = args["batch_size"]
else:
bs = 128
X_test = args["X_test"]
preds = iter_test(X_test).tolist()
new = []
for pred in preds:
new.append( np.eye(args["num_classes"])[pred].tolist() )
return new
if __name__ == '__main__':
x = ArffToArgs()
x.set_input("data/cpu_act.arff")
x.set_class_index("last")
x.set_impute(True)
x.set_binarize(True)
x.set_standardize(True)
x.set_arguments("adaptive=True;alpha=0.01;lambda=0;epochs=500;rmsprop=True")
args = x.get_args()
#args["debug"] = True
args["X_test"] = np.asarray(args["X_train"], dtype="float32")
model = train(args)
test(args, model)
if "expectation" in args:
new = []
for pred in preds:
label = int( round(expectation(pred)) )
new.append( np.eye(args["num_classes"])[label].tolist() )
return new
else:
return preds
if __name__ == '__main__':
x = ArffToArgs()
#x.set_input("data/auto_price.arff")
if len(sys.argv) != 3:
sys.argv.append("data/2dplanes.arff")
sys.argv.append("kappa")
x.set_input( sys.argv[1] )
print "Training on: %s" % sys.argv[1]
x.set_class_index("last")
x.set_impute(True)
x.set_binarize(True)
x.set_standardize(True)
if sys.argv[2] == "kappa":
#x.set_arguments("expectation=True;a=1;b=0;logistic=True;alpha=0.1;rmsprop=True;epochs=5000")
x.set_arguments("expectation=True;a=1;b=0;logistic=True;alpha=0.1;schedule=500;epochs=5000")
elif sys.argv[2] == "regression":
#x.set_arguments("regression=True;alpha=0.1;rmsprop=True;epochs=5000")
x.set_arguments("regression=True;alpha=0.1;schedule=500;epochs=5000")
else:
print "error!"
args = x.get_args()
args["debug"] = False
if b*bs >= len(filenames):
break
X_train_batch = get_batch(filenames, b*bs, (b+1)*bs)
#print (X_train_batch.shape)
#sys.stderr.write(" Batch #%i (%i-%i)\n" % ((b+1), (b*bs), ((b+1)*bs) ))
loss, loss_flat = iter_train( X_train_batch )
batch_train_losses.append(loss)
print (loss, loss_flat, sum(loss_flat > 0))
helper.plot_conv_activity( symbols.conv_layer, X_train_batch[1:2] )
sys.stderr.write( " train_loss = %f\n" % \
(np.mean(batch_train_losses)) )
current_weights = lasagne.layers.get_all_param_values(symbols.output_layer)
return (print_network(symbols.output_layer), current_weights)
if __name__ == '__main__':
f = ArffToArgs()
f.set_input("../mnist/mnist.meta.arff")
args = f.get_args()
f.close()
args["lambda"] = 0
args["alpha"] = 0.1
args["epochs"] = 10
args["dir"] = "../mnist/data"
weights = train(args)
X_test = np.asarray([load_image(x) for x in filenames], dtype="float32")
if "batch_size" in args:
bs = args["batch_size"]
else:
bs = 128
preds = iter_test(X_test).tolist()
return preds
if __name__ == '__main__':
f = ArffToArgs()
f.set_input("mnist.meta.arff")
args = f.get_args()
f.close()
args["lambda"] = 0
args["alpha"] = 0.01
args["epochs"] = 10
args["dir"] = "data"
weights = train(args)
args["X_test"] = args["X_train"]
preds = test(args, weights)
print("done!")
break
X_train_batch = get_batch(filenames, b * bs, (b + 1) * bs)
#print (X_train_batch.shape)
#sys.stderr.write(" Batch #%i (%i-%i)\n" % ((b+1), (b*bs), ((b+1)*bs) ))
loss, loss_flat = iter_train(X_train_batch)
batch_train_losses.append(loss)
print(loss, loss_flat, sum(loss_flat > 0))
helper.plot_conv_activity(symbols.conv_layer, X_train_batch[1:2])
sys.stderr.write( " train_loss = %f\n" % \
(np.mean(batch_train_losses)) )
current_weights = lasagne.layers.get_all_param_values(symbols.output_layer)
return (print_network(symbols.output_layer), current_weights)
if __name__ == '__main__':
f = ArffToArgs()
f.set_input("../mnist/mnist.meta.arff")
args = f.get_args()
f.close()
args["lambda"] = 0
args["alpha"] = 0.1
args["epochs"] = 10
args["dir"] = "../mnist/data"
weights = train(args)
from pyscript.pyscript import ArffToArgs
# test saving the pkl to an output file
f = ArffToArgs()
f.set_input("../datasets/iris.arff")
args = f.get_args()
f.save("iris.pkl.gz")
f.close()
# test normal
f = ArffToArgs()
f.set_input("../datasets/iris.arff")
args = f.get_args()
print f.output
f.close()
@uses(["dir"])
def train(args):
filenames = [ (args["dir"] + os.path.sep + elem) for elem in args["attr_values"]["filename"] ]
y_train = np.asarray(args["y_train"].flatten(), dtype="int32")
net1 = get_conv_net(filenames)
X_train = args["X_train"]
with Capturing() as output:
model = net1.fit(X_train, y_train)
return { "results": remove_colour("\n".join(output)),
"params": net1.get_all_params_values() }
def describe(args, model):
return model["results"]
def test(args, model):
filenames = [ (args["dir"] + os.path.sep + elem) for elem in args["attr_values"]["filename"] ]
net1 = get_conv_net(filenames)
net1.initialize()
net1.load_params_from(model["params"])
X_test = args["X_test"]
return net1.predict_proba(X_test).tolist()
if __name__ == "__main__":
f = ArffToArgs()
f.set_input("mnist.meta.arff")
args = f.get_args()
f.close()
args["dir"] = "data"
dd = train(args)
print(dd["results"])
new = []
for pred in preds:
label = int(round(expectation(pred)))
new.append(np.eye(args["num_classes"])[label].tolist())
return new
else:
return preds
if __name__ == '__main__':
x = ArffToArgs()
#x.set_input("data/auto_price.arff")
if len(sys.argv) != 3:
sys.argv.append("data/2dplanes.arff")
sys.argv.append("kappa")
x.set_input(sys.argv[1])
print "Training on: %s" % sys.argv[1]
x.set_class_index("last")
x.set_impute(True)
x.set_binarize(True)
x.set_standardize(True)
if sys.argv[2] == "kappa":
#x.set_arguments("expectation=True;a=1;b=0;logistic=True;alpha=0.1;rmsprop=True;epochs=5000")
x.set_arguments(
"expectation=True;a=1;b=0;logistic=True;alpha=0.1;schedule=500;epochs=5000"
)
elif sys.argv[2] == "regression":
#x.set_arguments("regression=True;alpha=0.1;rmsprop=True;epochs=5000")
x.set_arguments("regression=True;alpha=0.1;schedule=500;epochs=5000")
else:
print "error!"
