def test_t3(self):
# logger.info("Running Tests4Dataset1test1/test_t3")
ds = Dataset(TESTFILE2)
it0 = iter(ds.instances_original())
inst0 = next(it0)
indep, dep = inst0
# print("DEBUG: indep=", indep, file=sys.stderr)
assert indep == [['invincible', 'is', 'a', 'wonderful', 'movie', '.']]
assert dep == "pos"
# check low level conversion methods first
inst1 = [[['invincible', 'is', 'a', 'wonderful', 'movie', '.']], 'pos']
(indep1, dep1) = inst1
indep1_conv = ds.features(indep1)
logger.debug("Original indep1=%r", indep1)
logger.debug("Converted indep1=%r", indep1_conv)
ngram1 = indep1_conv[0]
assert len(ngram1) == 6
# print("DEBUG ngram1[0]=", ngram1[0], file=sys.stderr)
assert ngram1[0] == 3543
assert ngram1[1] == 9
it1 = iter(ds.instances_converted(train=False, convert=True))
rec = next(it1)
logger.debug("TESTFILE2 rec1=%r", rec)
logger.debug("TESTFILE2 info=%r" % ds.get_info())
(indep1_it, dep1_it) = rec
ngram1_it = indep1_it[0]
logger.debug("TESTFILE2 dep_it=%r", dep1_it)
# print("DEBUG dep1_it=", dep1_it, file=sys.stderr)
assert len(ngram1_it) == 6
assert ngram1_it[0] == 3543
assert ngram1_it[1] == 9
assert dep1_it == 1
def test_t4(self):
# logger.info("Running Tests4Dataset1test1/test_t4")
ds = Dataset(TESTFILE3)
logger.debug("TESTFILE3 attrs=%r",
ds.meta.get("featureInfo").get("attributes"))
# Features constructor finishes the vocab, so we need to re-initilize
features = ds.features
logger.debug("TESTFILE3 features=%r", features)
it1 = iter(ds.instances_original())
rec = next(it1)
logger.debug("TESTFILE3 rec1=%r", rec)
logger.debug("TESTFILE3 info=%r" % ds.get_info())
def test_t5(self):
# logger.info("Running Tests4Dataset1test1/test_t5")
ds = Dataset(TESTFILE4)
it1 = iter(ds.instances_converted(train=False, convert=True))
rec = next(it1)
indep, dep = rec
logger.debug("TESTFILE4: indep=%r" % indep)
logger.debug("TESTFILE4: dep=%r" % dep)
logger.debug("TESTFILE4 info=%r" % ds.get_info())
# the first row is a sequence of 3 elements, with 18 independent
# features and one of 17 different targets
# so we should convert this into 18 features which each now should have 3 values
# and 3 onehot vectors for the class
assert len(dep) == 3
assert len(indep) == 3 # 3 elements in the sequence
assert len(indep[0]) == 18
assert len(indep[1]) == 18
assert len(indep[2]) == 18
# check if the class is actually ADJ for all three targets
dep1 = dep[0]
dep2 = dep[1]
dep3 = dep[2]
t11 = ds.target.vocab.idx2string(dep1)
assert t11 == "ADJ"
t12 = ds.target.vocab.idx2string(dep2)
assert t12 == "ADJ"
t13 = ds.target.vocab.idx2string(dep3)
assert t13 == "ADJ"
# test getting batches in non-reshaped form
bit1 = ds.batches_converted(train=False,
convert=True,
batch_size=2,
reshape=False)
biter1 = iter(bit1)
batch1 = next(biter1)
# print("DEBUG: TESTFILE4 batch/noreshape=%s" % (batch1,), file=sys.stderr)
assert len(batch1) == 2
# test getting batches in reshaped form
bit2 = ds.batches_converted(train=False,
convert=True,
batch_size=2,
reshape=True)
biter2 = iter(bit2)
batch2 = next(biter2)
# print("DEBUG: TESTFILE4 batch/noreshape=%s" % (batch1,), file=sys.stderr)
bindep, bdep = batch2
assert len(bindep) == 18
assert len(bdep) == 2
assert len(bindep[0]) == 2
def test_t2(self):
# logger.info("Running Tests4Dataset1test1/test_t2")
ds = Dataset(TESTFILE1)
features = ds.features
s = features.size()
assert s == 34
it1 = iter(ds.instances_converted(train=False, convert=True))
rec = next(it1)
logger.debug("TESTFILE1 info=%r" % ds.get_info())
logger.debug("TESTFILE1 rec1=%r" % rec)
# we expect rec to be a pair: indep and dep
indep, dep = rec
# print("DEBUG: rec=", rec, file=sys.stderr)
# the indep part has as many values as there are features here
assert len(indep) == 34
# the dep part is the encoding for two nominal classes,
assert dep == 0