-
Notifications
You must be signed in to change notification settings - Fork 238
Expand file tree
/
Copy pathtest_ndarray.py
More file actions
279 lines (211 loc) · 8.74 KB
/
test_ndarray.py
File metadata and controls
279 lines (211 loc) · 8.74 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
import numpy as np
import orjson
import pytest
import torch
from pydantic.tools import parse_obj_as, schema_json_of
from docarray import BaseDoc
from docarray.base_doc.io.json import orjson_dumps
from docarray.typing import AudioNdArray, NdArray, TorchTensor
from docarray.typing.tensor import NdArrayEmbedding
from docarray.utils._internal.misc import is_tf_available
tf_available = is_tf_available()
if tf_available:
import tensorflow as tf
@pytest.mark.proto
def test_proto_tensor():
tensor = parse_obj_as(NdArray, np.zeros((3, 224, 224)))
tensor._to_node_protobuf()
def test_from_list():
tensor = parse_obj_as(NdArray, [[0.0, 0.0], [0.0, 0.0]])
assert (tensor == np.zeros((2, 2))).all()
def test_json_schema():
schema_json_of(NdArray)
def test_dump_json():
tensor = parse_obj_as(NdArray, np.zeros((3, 224, 224)))
orjson_dumps(tensor)
def test_load_json():
tensor = parse_obj_as(NdArray, np.zeros((2, 2)))
json = orjson_dumps(tensor)
print(json)
print(type(json))
new_tensor = orjson.loads(json)
assert (new_tensor == tensor).all()
def test_unwrap():
tensor = parse_obj_as(NdArray, np.zeros((3, 224, 224)))
ndarray = tensor.unwrap()
assert not isinstance(ndarray, NdArray)
assert isinstance(ndarray, np.ndarray)
assert isinstance(tensor, NdArray)
assert (ndarray == np.zeros((3, 224, 224))).all()
@pytest.mark.parametrize(
'tensor_class, tensor_type, tensor_fn',
[(NdArray, np.ndarray, np.zeros), (TorchTensor, torch.Tensor, torch.zeros)],
)
def test_ellipsis_in_shape(tensor_class, tensor_type, tensor_fn):
# ellipsis in the end, two extra dimensions needed
tensor = parse_obj_as(tensor_class[3, ...], tensor_fn((3, 128, 224)))
assert isinstance(tensor, tensor_class)
assert isinstance(tensor, tensor_type)
assert tensor.shape == (3, 128, 224)
# ellipsis in the middle, one extra dimension needed
tensor = parse_obj_as(tensor_class[3, ..., 224], tensor_fn((3, 128, 224)))
assert isinstance(tensor, tensor_class)
assert isinstance(tensor, tensor_type)
assert tensor.shape == (3, 128, 224)
# ellipsis in the beginning, two extra dimensions needed
tensor = parse_obj_as(tensor_class[..., 224], tensor_fn((3, 128, 224)))
assert isinstance(tensor, tensor_class)
assert isinstance(tensor, tensor_type)
assert tensor.shape == (3, 128, 224)
# more than one ellipsis in the shape
with pytest.raises(ValueError):
parse_obj_as(tensor_class[3, ..., 128, ...], tensor_fn((3, 128, 224)))
# bigger dimension than expected
with pytest.raises(ValueError):
parse_obj_as(tensor_class[3, 128, 224, ...], tensor_fn((3, 128)))
# no extra dimension needed
with pytest.raises(ValueError):
parse_obj_as(tensor_class[3, 128, 224, ...], tensor_fn((3, 128, 224)))
# wrong shape
with pytest.raises(ValueError):
parse_obj_as(tensor_class[3, 224, ...], tensor_fn((3, 128, 224)))
# passing only ellipsis as a shape
with pytest.raises(TypeError):
parse_obj_as(tensor_class[...], tensor_fn((3, 128, 224)))
@pytest.mark.parametrize(
'tensor_class, tensor_type, tensor_fn',
[(NdArray, np.ndarray, np.zeros), (TorchTensor, torch.Tensor, torch.zeros)],
)
def test_parametrized(tensor_class, tensor_type, tensor_fn):
# correct shape, single axis
tensor = parse_obj_as(tensor_class[128], tensor_fn(128))
assert isinstance(tensor, tensor_class)
assert isinstance(tensor, tensor_type)
assert tensor.shape == (128,)
# correct shape, multiple axis
tensor = parse_obj_as(tensor_class[3, 224, 224], tensor_fn((3, 224, 224)))
assert isinstance(tensor, tensor_class)
assert isinstance(tensor, tensor_type)
assert tensor.shape == (3, 224, 224)
# wrong but reshapable shape
tensor = parse_obj_as(tensor_class[3, 224, 224], tensor_fn((3, 224, 224)))
assert isinstance(tensor, tensor_class)
assert isinstance(tensor, tensor_type)
assert tensor.shape == (3, 224, 224)
# wrong and not reshapable shape
with pytest.raises(ValueError):
parse_obj_as(tensor_class[3, 224, 224], tensor_fn((224, 224)))
# test independent variable dimensions
tensor = parse_obj_as(tensor_class[3, 'x', 'y'], tensor_fn((3, 224, 224)))
assert isinstance(tensor, tensor_class)
assert isinstance(tensor, tensor_type)
assert tensor.shape == (3, 224, 224)
tensor = parse_obj_as(tensor_class[3, 'x', 'y'], tensor_fn((3, 60, 128)))
assert isinstance(tensor, tensor_class)
assert isinstance(tensor, tensor_type)
assert tensor.shape == (3, 60, 128)
with pytest.raises(ValueError):
parse_obj_as(tensor_class[3, 'x', 'y'], tensor_fn((4, 224, 224)))
with pytest.raises(ValueError):
parse_obj_as(tensor_class[3, 'x', 'y'], tensor_fn((100, 1)))
# test dependent variable dimensions
tensor = parse_obj_as(tensor_class[3, 'x', 'x'], tensor_fn((3, 224, 224)))
assert isinstance(tensor, tensor_class)
assert isinstance(tensor, tensor_type)
assert tensor.shape == (3, 224, 224)
with pytest.raises(ValueError):
tensor = parse_obj_as(tensor_class[3, 'x', 'x'], tensor_fn((3, 60, 128)))
with pytest.raises(ValueError):
tensor = parse_obj_as(tensor_class[3, 'x', 'x'], tensor_fn((3, 60)))
def test_np_embedding():
# correct shape
tensor = parse_obj_as(NdArrayEmbedding[128], np.zeros((128,)))
assert isinstance(tensor, NdArrayEmbedding)
assert isinstance(tensor, NdArray)
assert isinstance(tensor, np.ndarray)
assert tensor.shape == (128,)
# wrong shape at data setting time
with pytest.raises(ValueError):
parse_obj_as(NdArrayEmbedding[128], np.zeros((256,)))
# illegal shape at class creation time
with pytest.raises(ValueError):
parse_obj_as(NdArrayEmbedding[128, 128], np.zeros((128, 128)))
def test_parametrized_subclass():
c1 = NdArray[128]
c2 = NdArray[128]
assert issubclass(c1, c2)
assert issubclass(c1, NdArray)
assert issubclass(c1, np.ndarray)
assert not issubclass(c1, NdArray[256])
def test_parametrized_instance():
t = parse_obj_as(NdArray[128], np.zeros(128))
assert isinstance(t, NdArray[128])
assert isinstance(t, NdArray)
assert isinstance(t, np.ndarray)
assert not isinstance(t, NdArray[256])
assert not isinstance(t, NdArray[2, 64])
assert not isinstance(t, NdArray[2, 2, 32])
def test_parametrized_equality():
t1 = parse_obj_as(NdArray[128], np.zeros(128))
t2 = parse_obj_as(NdArray[128], np.zeros(128))
t3 = parse_obj_as(NdArray[128], np.ones(128))
assert (t1 == t2).all()
assert not (t1 == t3).any()
def test_parametrized_operations():
t1 = parse_obj_as(NdArray[128], np.zeros(128))
t2 = parse_obj_as(NdArray[128], np.zeros(128))
t_result = t1 + t2
assert isinstance(t_result, np.ndarray)
assert isinstance(t_result, NdArray)
assert isinstance(t_result, NdArray[128])
def test_class_equality():
assert NdArray == NdArray
assert NdArray[128] == NdArray[128]
assert NdArray[128] != NdArray[256]
assert NdArray[128] != NdArray[2, 64]
assert not NdArray[128] == NdArray[2, 64]
assert NdArrayEmbedding == NdArrayEmbedding
assert NdArrayEmbedding[128] == NdArrayEmbedding[128]
assert NdArrayEmbedding[128] != NdArrayEmbedding[256]
assert AudioNdArray == AudioNdArray
assert AudioNdArray[128] == AudioNdArray[128]
assert AudioNdArray[128] != AudioNdArray[256]
def test_class_hash():
assert hash(NdArray) == hash(NdArray)
assert hash(NdArray[128]) == hash(NdArray[128])
assert hash(NdArray[128]) != hash(NdArray[256])
assert hash(NdArray[128]) != hash(NdArray[2, 64])
assert not hash(NdArray[128]) == hash(NdArray[2, 64])
assert hash(NdArrayEmbedding) == hash(NdArrayEmbedding)
assert hash(NdArrayEmbedding[128]) == hash(NdArrayEmbedding[128])
assert hash(NdArrayEmbedding[128]) != hash(NdArrayEmbedding[256])
assert hash(AudioNdArray) == hash(AudioNdArray)
assert hash(AudioNdArray[128]) == hash(AudioNdArray[128])
assert hash(AudioNdArray[128]) != hash(AudioNdArray[256])
@pytest.mark.parametrize(
'tensor',
[
torch.zeros(10),
TorchTensor(torch.zeros(10)),
np.zeros(10),
],
)
def test_torch_numpy_to_ndarray(tensor):
class MyAudioDoc(BaseDoc):
tensor: NdArray
doc = MyAudioDoc(tensor=tensor)
assert isinstance(doc.tensor, np.ndarray)
assert isinstance(doc.tensor, NdArray)
assert isinstance(doc.tensor, NdArray[10])
@pytest.mark.tensorflow
def test_tensorflow_to_ndarray():
class MyAudioDoc(BaseDoc):
tensor: NdArray
doc = MyAudioDoc(
tensor=tf.zeros(
10,
)
)
assert isinstance(doc.tensor, np.ndarray)
assert isinstance(doc.tensor, NdArray)
assert isinstance(doc.tensor, NdArray[10])