#Load the preprocessed version, saving on computation
X = np.load('../data/data_cmu.npz')['clips']
X = np.swapaxes(X, 1, 2).astype(theano.config.floatX)
X = X[:,:-4]
preprocess = np.load('../data/Joe/preprocess.npz')
X = (X - preprocess['Xmean']) / preprocess['Xstd']
H = np.load('../data/Joe/HiddenActivations.npz')['Orig']
from network import network
network.load([
None,
'../models/conv_ae/layer_0.npz', None, None,
'../models/conv_ae/layer_1.npz', None, None,
'../models/conv_ae/layer_2.npz', None, None,
])
for layer in network.layers:
if isinstance(layer, NoiseLayer): layer.amount = 0.0
if isinstance(layer, Pool1DLayer): layer.depooler = lambda x, **kw: x/2
# Go through inputs 1by1
for input in range(len(X)):
Xorig = X[input:input+1]
#Theano shared object to pass to network
shared = theano.shared(H[input:input+1])
X = np.load('../data/data_cmu.npz')['clips']
X = np.swapaxes(X, 1, 2).astype(theano.config.floatX)
X = X[:, :-4]
preprocess = np.load('../data/Joe/preprocess.npz')
X = (X - preprocess['Xmean']) / preprocess['Xstd']
from network import network
network.load([
None,
'../models/conv_ae/layer_0.npz',
None,
None,
'../models/conv_ae/layer_1.npz',
None,
None,
'../models/conv_ae/layer_2.npz',
None,
None,
])
for layer in network.layers:
if isinstance(layer, NoiseLayer): layer.amount = 0.0
if isinstance(layer, Pool1DLayer): layer.depooler = lambda x, **kw: x / 2
while True:
index = rng.randint(len(X) - 1)
amount = 0.5
def test(X,y):
global model
accuracies = []
costs = []
for i in range(0,X.shape[0],64):
end = min(i+64,X.shape[0])
model.test(X[i:end],y[i:end],accuracies,costs)
accuracies = np.array(accuracies)
costs = np.array(costs)
print("Accuracy : ",np.mean(accuracies))
print("Cost : ",np.sum(costs))
if __name__ == "__main__":
global model
model = network.load("model.json")
data = json.load(open("data/data.json","rb"))
trainX = np.array(data['trainX'])
trainY = np.array(data['trainY'],dtype=np.int32)
validX = np.array(data['validX'])
validY = np.array(data['validY'],dtype=np.int32)
testX = np.array(data['testX'])
testY = np.array(data['testY'],dtype=np.int32)
print("TRAIN SET")
test(trainX,trainY)
print("\n\nVALIDATION SET")
test(validX,validY)
print("\n\nTEST SET")