import numpy
from pylearn2_timit.timitlpc import TIMITlpc
from pylearn2.space import CompositeSpace, VectorSpace, IndexSpace
from pylearn2.format.target_format import OneHotFormatter
valid = TIMITlpc("valid", frame_length=160, overlap=159, start=10, stop=11)
valid._iter_data_specs = (CompositeSpace((IndexSpace(dim=3,max_labels=61), VectorSpace(dim=10),)), ('phones', 'lpc_features'))
formatter = OneHotFormatter(max_labels=62)
f = lambda x: formatter.format(numpy.asarray(x, dtype=int), mode='merge')
#valid._iter_convert = [f, None]
it = valid.iterator(mode='random_uniform', batch_size=100, num_batches=100)
from pylearn2.termination_criteria import EpochCounter
from pylearn2.training_algorithms.sgd import SGD
from pylearn2.training_algorithms import learning_rule
from pylearn2.train import Train
from pylearn2.train_extensions import best_params
from pylearn2.costs.cost import SumOfCosts
from pylearn2.costs.mlp import *
from pylearn2.space import *
import cPickle as pickle
import theano
from sys import argv
frame_len = 160
overlap = 0
train = TIMITlpc("train", frame_len, overlap, start=0, stop=100)
valid = TIMITlpc("valid", frame_len, overlap, start=0, stop=10)
test = TIMITlpc("test", frame_len, overlap, start=0, stop=50)
formatter = OneHotFormatter(max_labels=62)
f = lambda x: formatter.format(np.asarray(x, dtype=int), mode='merge')
iter_convert = [None, f]
train._iter_convert = iter_convert
valid._iter_convert = iter_convert
test._iter_convert = iter_convert
# Model used to predict the next sample. This model is never trained,
# as it uses parameters predicted by the next model.
i0 = VectorSpace(3*62)
