def define_stackrnn(LEN, DIM_IN, DIM_OUT):
model = StackedRNN(
timespan=LEN,
input_dim=DIM_IN,
output_dim=DIM_OUT,
cells=[512, 512, 512],
)
model.build_generator('temp/e-comp_simple_rnn.h5')
# model.build_generator('temp/temp.h5')
return model
def define_stackrnn_hist(LEN, DIM_IN, DIM_OUT):
model = StackedRNN(
timespan=LEN,
input_dim=DIM_IN,
output_dim=DIM_OUT,
cells=[512, 512, 512],
)
# model.build_generator('temp/stackedrnn_hist.h5')
model.build_generator('temp/stackedrnn_hist_patch_T=0.05.h5')
return model
def test_stackrnn(data, weight='temp/simple_rnn.h5'):
model = StackedRNN(timespan=LEN,
input_dim=DIM,
output_dim=DIM,
cells=[512, 512, 512])
model.build_generator(weight)
moving_mean = []
for _ in range(TEST):
ind = RNG.randint(len(data))
start = RNG.randint(data[ind].shape[0] - LEN - 1)
x = np.array(data[ind][start:start + LEN])
x = to_categorical(x.flatten(), num_classes=DIM).\
reshape(x.shape[0], DIM)
score = bleu(model=model,
prior=x,
prefix_length=PREFIX,
tokenize=tokenize)
moving_mean.append(score)
print np.mean(moving_mean)
def test_emb_stackrnn(data, weight='temp/emb_stackrnn.h5'):
emb_w = load_model('temp/emb.h5').layers[2].layers[1].get_weights()
model = StackedRNN(timespan=LEN,
input_dim=DIM,
output_dim=DIM,
cells=[512, 512, 512],
embedding=Embedding(DIM, 512, trainable=False),
embedding_w=emb_w)
model.build_generator(weight)
moving_mean = []
for _ in range(TEST):
ind = RNG.randint(len(data))
start = RNG.randint(data[ind].shape[0] - LEN - 1)
x = np.array(data[ind][start:start + LEN])
score = bleu(model=model,
prior=x,
prefix_length=PREFIX,
tokenize=tokenize)
moving_mean.append(score)
print np.mean(moving_mean)
start = np.random.randint(data[ind].shape[0] - LEN - 1)
x.append(data[ind][start:start + LEN])
y.append(data[ind][start + 1:start + LEN + 1])
x = np.array(x)
y = np.array(y)
# sparse to complex
x = to_categorical(x.flatten(), num_classes=DIM).\
reshape(x.shape[0], x.shape[1], DIM)
y = y.reshape(x.shape[0], x.shape[1], 1)
yield (x, y)
# model
model = StackedRNN(
timespan=LEN,
input_dim=DIM,
output_dim=DIM,
cells=[512, 512, 512],
)
# block_kwargs={'kernel_regularizer': l2(1e-5)})
model.build()
# model.model.load_weights('temp/stackedrnn_kernel_l2.h5')
# model.model.load_weights('temp/simple_rnn.h5')
# model.train(data_generator(),
# opt=1e-5,
# steps_per_epoch=30,
# epochs=100,
# save_path='temp/simple_rnn.h5')
model.build_generator('temp/e-comp_simple_rnn.h5')
res = model.generate(seed=32, length=2000)
ind = np.random.randint(len(data))
start = np.random.randint(data[ind].shape[0] - LEN - 1)
x.append(data[ind][start:start + LEN])
y.append(data[ind][start + 1:start + LEN + 1])
x = np.array(x)
y = np.array(y)
# sparse to complex
y = y.reshape(x.shape[0], x.shape[1], 1)
yield (x, y)
# model
emb_w = load_model('temp/emb.h5').layers[2].layers[1].get_weights()
model = StackedRNN(timespan=LEN,
input_dim=DIM,
output_dim=DIM,
cells=[512, 512, 512],
embedding=Embedding(DIM, 512, trainable=False),
embedding_w=emb_w)
model.build()
model.model.load_weights('temp/emb_stackrnn.h5')
model.train(data_generator(),
opt=1e-5,
steps_per_epoch=20,
epochs=200,
save_path='temp/emb_stackrnn.h5')
res = model.generate('temp/emb_stackrnn.h5', seed=64, length=5000)
mid = Song()
track = mid.add_track()
for msgi in AllInOneCoder().decode(res):
print '| Layer[{}]:'.format(ind), cell.max(), cell.min(),
print
# self.keyboard = self.keyboard*0.9
note = note.argmax()
if note <= 127:
self.keyboard[0, note] = 1
elif note <= 128 + 127:
self.keyboard[0, note - 128] = 0
self.imshows[-1].set_data(self.keyboard)
self.canvas[-1].draw()
if __name__ == '__main__':
model = StackedRNN(timespan=LEN,
input_dim=DIM,
output_dim=DIM,
cells=[512, 512, 512])
model.build()
# generator = model.build_generator('temp/simple_rnn.h5')
generator = model.build_generator('temp/stackedrnn_act_l1.h5')
lstm_layers = []
for layer in generator.layers:
if isinstance(layer, LSTM):
lstm_layers.append(layer)
print lstm_layers
window = VisualizeWindow(model=model, lstm_layers=lstm_layers)
window.connect("delete-event", Gtk.main_quit)
window.show_all()
Gtk.main()
ind = np.random.randint(len(data))
start = np.random.randint(data[ind].shape[0]-LEN-1)
x.append(data[ind][start:start+LEN])
y.append(data[ind][start+1:start+LEN+1])
x = np.array(x)
y = np.array(y)
# sparse to complex
x = to_categorical(x.flatten(), num_classes=DIM).\
reshape(x.shape[0], x.shape[1], DIM)
y = y.reshape(x.shape[0], x.shape[1], 1)
yield (x, y)
# model
model = StackedRNN(timespan=LEN,
input_dim=DIM,
output_dim=DIM,
cells=[128, 128, 128],)
model.build()
model.model.load_weights('temp/start_with_me.h5')
# model.train(data_generator(),
# opt=optimizers.Adam(1e-3),
# steps_per_epoch=30,
# epochs=300,
# save_path='temp/start_with_me.h5')
model.build_generator('temp/start_with_me.h5')
res = model.generate(seed=32, length=2000, verbose=0)
# store
mid = Song()
track = mid.add_track()
h.append(norm(h_, H_TEMP))
x = np.array(x)
y = np.array(y)
h = np.array(h)
# sparse to complex
x = to_categorical(x.flatten(), num_classes=DIM_IN).\
reshape(x.shape[0], x.shape[1], DIM_IN)
x[:, :, -len(NOTE_NUMBER):] = h[:, None, :]
y = y.reshape(x.shape[0], x.shape[1], 1)
yield (x, y)
# model
model = StackedRNN(
timespan=LEN,
input_dim=DIM_IN,
output_dim=DIM_OUT,
cells=[512, 512, 512],
)
model.build()
# model.model.load_weights('temp/stackedrnn_hist.h5')
# model.train(data_generator(),
# opt=8e-4,
# steps_per_epoch=30,
# epochs=300,
# save_path='temp/stackedrnn_hist.h5')
C_scale = np.array([2, 0, 1, 0, 1, 1, 0, 1, 0, 1, 0, 1])
model.build_generator('temp/stackedrnn_hist.h5')
res = model.generate(seed=64, length=3000, addition=norm(C_scale, ))
events = [
h.append(norm(h_, H_TEMP))
x = np.array(x)
y = np.array(y)
h = np.array(h)
# sparse to complex
x = to_categorical(x.flatten(), num_classes=DIM_IN).\
reshape(x.shape[0], x.shape[1], DIM_IN)
x[:, :, -len(NOTE_NUMBER):] = h[:, None, :]
y = y.reshape(x.shape[0], x.shape[1], 1)
yield (x, y)
# model
model = StackedRNN(
timespan=LEN,
input_dim=DIM_IN,
output_dim=DIM_OUT,
cells=[512, 512, 512],
)
model.build()
# model.model.load_weights('temp/stackedrnn_hist_easy.h5')
# model.train(data_generator(),
# opt=1e-3,
# steps_per_epoch=30,
# epochs=300,
# save_path='temp/stackedrnn_hist_easy.h5')
model.build_generator('temp/stackedrnn_hist_easy.h5')
res = model.generate(seed=64,
length=3000,
addition=norm(hist[1], H_TEMP),
verbose=0)