model.add(RepeatVector(DIGITS + 1))
# The decoder RNN could be multiple layers stacked or a single keras_layer
for _ in range(LAYERS):
model.add(RNN(HIDDEN_SIZE, return_sequences=True))
# For each of step of the output sequence, decide which character should be chosen
model.add(TimeDistributedDense(convertor.get_dim()))
model.add(Activation("softmax"))
model.compile(loss="categorical_crossentropy", optimizer="adam")
# Train the model each generation and show predictions against the validation dataset
for iteration in range(1, 200):
print()
print("-" * 50)
print("Iteration", iteration)
model.fit(X_train, y_train, batch_size=BATCH_SIZE, nb_epoch=1, validation_data=(X_val, y_val), show_accuracy=True)
###
# Select 10 samples from the validation set at random so we can visualize errors
for i in range(10):
ind = np.random.randint(0, len(X_val))
rowX, rowy = X_val[np.array([ind])], y_val[np.array([ind])]
preds = model.predict_classes(rowX, verbose=0)
q = convertor.decode(rowX[0], invert=True)
correct = convertor.decode(rowy[0])
guess = convertor.decode(preds[0], calc_argmax=False)
print("Q", q)
print("T", correct)
print(colors.ok + "☑" + colors.close if correct == guess else colors.fail + "☒" + colors.close, guess)
print("---")
nb_epoch=1,
show_accuracy=True,
verbose=2,
validation_data={
'X': D_X_val,
'y': D_y_val
})
###
# Select 10 samples from the validation set at random so we can visualize errors
for i in range(10):
ind = np.random.randint(0, len(D_X_val))
rowX, rowy = D_X_val[np.array([ind])], D_y_val[np.array([ind])]
preds = model.predict({'X': rowX},
class_mode={'y': "categorical"},
verbose=0)['y']
q = convertor.decode(rowX[0], invert=INVERT)
correct = convertor.decode(rowy[0])
guess = convertor.decode(preds[0], calc_argmax=False)
print('Q', q)
print('T', correct)
print(
colors.ok + '☑' +
colors.close if correct == guess else colors.fail + '☒' +
colors.close, guess)
print('---')
import os
usr_home = os.path.expanduser('~')
save_dir = os.path.join(usr_home, ".dlx/saved_model/")
if not os.path.isdir(save_dir):
os.mkdir(save_dir)
validation_data={
'X': D_X_val,
'y': D_y_val
})
###
# Select 10 samples from the validation set at random so we can visualize errors
for i in range(10):
ind = np.random.randint(0, len(D_X_val))
rowX, rowy = D_X_val[np.array([ind])], D_y_val[np.array([ind])]
preds = model.predict({'X': rowX},
class_mode={
'y': "categorical",
'alpha': None
},
verbose=0)
q = convertor.decode(rowX[0], invert=True)
correct = convertor.decode(rowy[0])
guess = convertor.decode(preds['y'][0], calc_argmax=False)
alpha = preds['alpha']
print('Q', q)
print('T', correct)
print(
colors.ok + '☑' +
colors.close if correct == guess else colors.fail + '☒' +
colors.close, guess)
print('Alpha:', alpha[0].shape)
sys.stdout.write(' A\\Q')
for j in range(len(q)):
sys.stdout.write("%6s" % q[j])
###
# Select 10 samples from the validation set at random so we can visualize errors
for ii in range(10):
ind = np.random.randint(0, len(D_A_val))
rowA, rowB, rowy = D_A_val[np.array([ind])], D_B_val[np.array(
[ind])], D_y_val[np.array([ind])]
preds = model.predict({
'A': rowA,
'B': rowB
},
class_mode={
'y': "categorical",
'alpha': None
},
verbose=0)
string_A = convertor.decode(rowA[0], invert=True, index=True)
string_B = convertor.decode(rowB[0], invert=True, index=True)
correct = convertor.decode(rowy[0], invert=True)
guess = convertor.decode(preds['y'][0], calc_argmax=False)[::-1]
alpha = preds['alpha']
print('A', string_A)
print(colors.ok + '+' + colors.close)
print('B', string_B)
print(colors.ok + '=' + colors.close)
print('T', correct)
print(
colors.ok + '☑' +
colors.close if correct == guess else colors.fail + '☒' +
colors.close, guess)
history = model.fit(
data={"A": D_A_train, "B": D_B_train, "y": D_y_train},
batch_size=BATCH_SIZE,
nb_epoch=1,
show_accuracy=True,
verbose=2,
validation_data={"A": D_A_val, "B": D_B_val, "y": D_y_val},
)
###
# Select 10 samples from the validation set at random so we can visualize errors
for ii in range(10):
ind = np.random.randint(0, len(D_A_val))
rowA, rowB, rowy = D_A_val[np.array([ind])], D_B_val[np.array([ind])], D_y_val[np.array([ind])]
preds = model.predict({"A": rowA, "B": rowB}, class_mode={"y": "categorical", "alpha": None}, verbose=0)
string_A = convertor.decode(rowA[0], invert=True, index=True)
string_B = convertor.decode(rowB[0], invert=True, index=True)
correct = convertor.decode(rowy[0], invert=True)
guess = convertor.decode(preds["y"][0], calc_argmax=False)[::-1]
alpha = preds["alpha"]
print("A", string_A)
print(colors.ok + "+" + colors.close)
print("B", string_B)
print(colors.ok + "=" + colors.close)
print("T", correct)
print(colors.ok + "☑" + colors.close if correct == guess else colors.fail + "☒" + colors.close, guess)
if ii == 9:
print("Alpha:", alpha[0].shape)
sys.stdout.write("GUESS A\\B")