def test_simple_long(lstm_kwargs_1):
model = lstm.LSTMModel('test-simple-long',
2,
use_long=True,
**lstm_kwargs_1)
model.train([[1, 0] * 6000])
s = model.sample([1])
assert isinstance(s, list)
print(s)
model.analyze([1, 0])
model.save_to_file()
model.sample([1])
model = lstm.LSTMModelFromFile('test-simple-long', **lstm_kwargs_1)
s = model.sample([1])
assert isinstance(s, list)
print(s)
model.analyze([1, 0])
model.train([[1, 0] * 6000])
model.sample([1])
model.analyze([1, 0])
model.save_to_file()
model.train([[1, 0]])
model.train([[1, 0] * 6000], count=True)
def test_simple_long_skip_padding(lstm_kwargs_1):
model = lstm.LSTMModel('test-simple-long',
2,
use_long=True,
skip_padding=True,
**lstm_kwargs_1)
model.train([[1, 0] * 1000])
model.train([[1, 0]])
model.save_to_file()
model = lstm.LSTMModelFromFile('test-simple-long', **lstm_kwargs_1)
model.train([[1, 0] * 1000])
model.train([[1, 0]])
model.save_to_file()
def test_simple_short(lstm_kwargs_1, lstm_kwargs_2):
model = lstm.LSTMModel('test-simple', 2, **lstm_kwargs_1)
model.train([[1, 0]] * 1000)
s = model.sample()
assert isinstance(s, list)
print(s)
model.analyze([1, 0])
model.save_to_file()
model.sample([1])
model = lstm.LSTMModelFromFile('test-simple', **lstm_kwargs_1)
s = model.sample()
assert isinstance(s, list)
print(s)
model.analyze([1, 0])
model.train([[1, 0]] * 1000)
model.sample()
model.analyze([1, 0])
model.save_to_file()
# Double saving is ok
model.save_to_file()
model = lstm.LSTMModel('test-simple', 2, **lstm_kwargs_1)
model.train([[1, 0]] * 1000, autosave=30)
model = lstm.LSTMModel('test-simple', 2, **lstm_kwargs_2)
model.train([[1, 0]] * 1000, autosave=30)
model.train([[1, 0]] * 1000, autosave=True)
model = lstm.LSTMModel('test-simple', 2, **lstm_kwargs_2)
model.train([[1, 0]] * 1000, autosave=30, count=True)
model = lstm.LSTMModelFromFile('test-simple-copy',
from_name='test-simple',
training_steps=100,
**lstm_kwargs_1)
model.train([[1, 0]] * 1000, autosave=30)
#!/usr/bin/env python3
import setup
from models import lstm
model = lstm.LSTMModel('simple', 2, tag=setup.tag)
model.train([[1, 0]] * 10000)
model.save_to_file()
#!/usr/bin/env python3
import setup
from models import lstm
import numpy as np
model = lstm.LSTMModel('count-1-0', 2, tag=setup.tag)
data = [(lambda n: [1] * n + [0] * n)(np.random.randint(0, 10))
for i in range(100000)]
model.train(data, count=True, autosave=True)
model.save_to_file()
#!/usr/bin/env python3
import setup
from models import lstm
# Use skip_padding to remove the last training example which is padded
# to prevent spikes in loss_max
model = lstm.LSTMModel('simple-long', 2, use_long=True, skip_padding=True, tag=setup.tag)
data = [[1, 0] * 100000]
model.train(data, autosave=300, cont=True)