def info(ses, fn, t, prt, kind):
class Opt:
def __init__(self, t):
self.after = t
self.before = float('inf')
self.overview = 'none'
# xxx assumes this exists
time_metric = util.join('serverStatus', 'localTime')
# find and print the first sample after t (only)
for metrics in read(ses, fn, Opt(t), progress=False):
if kind=='raw':
for sample, sample_time in enumerate(metrics[time_metric]):
sample_time = sample_time / 1000.0
if sample_time >= t:
break
prt('%s at t=%.3f (%s)' % (fn, t, util.f2s(t)))
util.print_sample(metrics, sample, prt)
break
elif kind=='metadata':
prt('metadata at t=%.3f (%s)' % (t, util.f2s(t)))
if metrics.metadata:
util.print_bson_doc(metrics.metadata, prt, ' ')
else:
prt(' NOT AVAILABLE')
break
def info(ses, fn, t, prt, kind):
class Opt:
def __init__(self, t):
self.after = t
self.before = float('inf')
self.overview = 'none'
# xxx assumes this exists
time_metric = util.join('serverStatus', 'localTime')
# find and print the first sample after t (only)
for metrics in read(ses, fn, Opt(t), progress=False):
if kind == 'raw':
for sample, sample_time in enumerate(metrics[time_metric]):
sample_time = sample_time / 1000.0
if sample_time >= t:
break
prt('%s at t=%.3f (%s)' % (fn, t, util.f2s(t)))
util.print_sample(metrics, sample, prt)
break
elif kind == 'metadata':
prt('metadata at t=%.3f (%s)' % (t, util.f2s(t)))
if metrics.metadata:
util.print_bson_doc(metrics.metadata, prt, ' ')
else:
prt(' NOT AVAILABLE')
break
def model(data,
ix_to_char,
char_to_ix,
vocab_size,
num_iterations=35000,
n_a=50):
"""
Trains the model and generates jokes.
Arguments:
data -- text corpus
ix_to_char -- dictionary that maps the index to a character
char_to_ix -- dictionary that maps a character to an index
num_iterations -- number of iterations to train the model for
n_a -- number of units of the RNN cell
vocab_size -- number of unique characters in vocabulary
Returns:
parameters -- learned parameters
"""
n_x, n_y = vocab_size, vocab_size
parameters = initialize_parameters(n_a, n_x, n_y)
loss = get_initial_loss(vocab_size, 10)
# Shuffle list of all jokes
np.random.seed(0)
np.random.shuffle(data)
# Initialize the hidden state
a_prev = np.zeros((n_a, 1))
# Optimization loop
for j in range(num_iterations):
index = j % len(data)
X = [None] + [char_to_ix[ch] for ch in data[index]]
Y = X[1:] + [char_to_ix["\n"]]
curr_loss, gradients, a_prev = optimize(X,
Y,
a_prev,
parameters,
vocab_size,
learning_rate=0.01)
# Use a latency trick to keep the loss smooth.
loss = smooth(loss, curr_loss)
# Every 1000 iterations, generate n characters to check if the model is
# learning properly
if j % 1000 == 0:
print('Iteration: %d, Loss: %f' % (j, loss) + '\n')
for name in range(10):
sampled_indices = sample(parameters, char_to_ix)
print_sample(sampled_indices, ix_to_char)
print('\n')
return parameters
def info(cls, ses, fn, t, kind):
if kind != 'raw':
return
def putln(*s):
ses.put(' '.join(str(ss) for ss in s) + '\n')
file = cls.get(fn)
for chunk in file.chunks:
if t >= chunk[file.time_key][0]:
for sample, sample_time in enumerate(chunk[file.time_key]):
if sample_time >= t:
putln('%s at t=%.3f (%s)' % (fn, t, util.f2s(t)))
util.print_sample(chunk, sample, putln)
return
def info(cls, ses, fn, t, kind):
if kind != 'raw':
return
def putln(*s):
ses.put(' '.join(str(ss) for ss in s) + '\n')
file = cls.get(fn)
for chunk in file.chunks:
if t >= chunk[file.time_key][0]:
for sample, sample_time in enumerate(chunk[file.time_key]):
if sample_time >= t:
putln('%s at t=%.3f (%s)' % (fn, t, util.f2s(t)))
util.print_sample(chunk, sample, putln)
return
