def transformType(self, cln, values, tpe):
if tpe == "as_is":
return values
cd = coders.get_coder(tpe, values, self)
self.coders[cln] = cd
if isinstance(cd, coders.ImageCoder):
self.imageCoders.append(cd)
return cd
def test_binary_bool(self):
a=np.array([True,False,True,False])
bc=coders.get_coder("binary",a, None)
self.assertEqual(bc[0], 1, "should be zero")
self.assertEqual(bc[1], 0, "should be one")
v=bc._decode(np.array([0.6]))
self.assertEqual(v, True, "should be one")
v=bc._decode(np.array([0.2]))
self.assertEqual(v, False, "should be zero")
pass
def test_binary_str2(self):
a=np.array(["","1","","1"])
bc=coders.get_coder("binary",a, None)
self.assertEqual(bc[0], 0, "should be zero")
self.assertEqual(bc[1], 1, "should be one")
v=bc._decode(np.array([0.6]))
self.assertEqual(v, "1", "should be one")
v=bc._decode(np.array([0.2]))
self.assertEqual(v, "", "should be zero")
pass
def test_categorical_pd(self):
a=np.array([math.nan,1,2,1])
bc=coders.get_coder("categorical_one_hot",a, None)
self.assertEqual(bc[0][2], True, "should be zero")
self.assertEqual(bc[0][1], False, "should be one")
v=bc._decode(np.array([0.3,0.4,0.45]))
self.assertEqual(math.isnan(v),True, "should be one")
v=bc._decode(np.array([0.2,0.1,0.1]))
self.assertEqual(v, 1, "should be zero")
pass
def test_categorical_str2(self):
a=np.array(["","b","c","b"])
bc=coders.get_coder("categorical_one_hot",a, None)
self.assertEqual(bc[0][0], True, "should be zero")
self.assertEqual(bc[0][1], False, "should be one")
v=bc._decode(np.array([0.3,0.4,0.45]))
self.assertEqual(v, "c", "should be one")
v=bc._decode(np.array([0.2,0.1,0.1]))
self.assertEqual(v, "", "should be zero")
pass
def test_multiclass2(self):
a=np.array(["1","","",""])
bc=coders.get_coder("multi_class",a, None)
val=bc[0]
self.assertEqual((val==np.array([True])).sum(), 1,"Fixing format")
for i in range(len(a)):
val=bc[i]
r=bc._decode(val)
self.assertEqual(r, a[i], "Decoding should work also")
pass