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
|
from typing import List, Optional
import numpy
import pytest
import srsly
from numpy.testing import assert_almost_equal
from thinc.api import Dropout, Model, NumpyOps, registry, with_padded
from thinc.backends import NumpyOps
from thinc.compat import has_torch
from thinc.types import Array2d, Floats2d, FloatsXd, Padded, Ragged, Shape
from thinc.util import data_validation, get_width
OPS = NumpyOps()
class NoDropoutOps(NumpyOps):
def get_dropout_mask(self, shape: Shape, drop: Optional[float]) -> FloatsXd:
if drop is None or drop <= 0:
return self.xp.ones(shape, dtype="f")
else:
raise ValueError("During prediction, dropout should not be applied")
array1d = OPS.xp.asarray([1, 2, 3], dtype="f")
array1dint = OPS.xp.asarray([1, 2, 3], dtype="i")
array2d = OPS.xp.asarray([[4, 2, 3, 4], [1, 5, 3, 1], [9, 8, 5, 7]], dtype="f")
array2dint = OPS.xp.asarray([[1, 2, 3], [4, 5, 6]], dtype="i")
array3d = OPS.xp.zeros((3, 3, 3), dtype="f")
ragged = Ragged(array2d, OPS.xp.asarray([2, 1], dtype="i"))
padded = Padded(
array3d, array1d, OPS.asarray1i([1, 2, 3, 4]), OPS.asarray1i([1, 2, 3, 4])
)
width = array2d.shape[1]
vectors = numpy.zeros((array2dint.max(), 1), dtype="f")
def assert_data_match(Y, out_data):
assert type(Y) == type(out_data)
if isinstance(out_data, OPS.xp.ndarray):
assert isinstance(Y, OPS.xp.ndarray)
assert out_data.ndim == Y.ndim
elif isinstance(out_data, Ragged):
assert isinstance(Y, Ragged)
assert out_data.data.ndim == Y.data.ndim
assert out_data.lengths.ndim == Y.lengths.ndim
elif isinstance(out_data, Padded):
assert isinstance(Y, Padded)
assert out_data.data.ndim == Y.data.ndim
assert out_data.size_at_t.ndim == Y.size_at_t.ndim
assert len(out_data.lengths) == len(Y.lengths)
assert len(out_data.indices) == len(Y.indices)
elif isinstance(out_data, (list, tuple)):
assert isinstance(Y, (list, tuple))
assert all(isinstance(x, numpy.ndarray) for x in Y)
else:
pytest.fail(f"wrong output of {type(Y)}: {Y}")
TEST_CASES_SUMMABLE = [
# Array to array
("Dish.v1", {}, array2d, array2d),
("Dish.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("Dropout.v1", {}, array2d, array2d),
("LayerNorm.v1", {}, array2d, array2d),
("Linear.v1", {}, array2d, array2d),
("Logistic.v1", {}, array2d, array2d),
("Maxout.v1", {}, array2d, array2d),
("Maxout.v1", {"normalize": True, "dropout": 0.2}, array2d, array2d),
("Maxout.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("Mish.v1", {}, array2d, array2d),
("Mish.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("Mish.v1", {"normalize": True, "dropout": 0.2}, array2d, array2d),
("Relu.v1", {}, array2d, array2d),
("Relu.v1", {"normalize": True, "dropout": 0.2}, array2d, array2d),
("Sigmoid.v1", {}, array2d, array2d),
("Sigmoid.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("ClippedLinear.v1", {}, array2d, array2d),
("ClippedLinear.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("ReluK.v1", {}, array2d, array2d),
("ReluK.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("HardSigmoid.v1", {}, array2d, array2d),
("HardSigmoid.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("HardTanh.v1", {}, array2d, array2d),
("HardTanh.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("HardSwish.v1", {}, array2d, array2d),
("HardSwish.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("HardSwishMobilenet.v1", {}, array2d, array2d),
("HardSwishMobilenet.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("Swish.v1", {}, array2d, array2d),
("Swish.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("Gelu.v1", {}, array2d, array2d),
("Gelu.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("sigmoid_activation.v1", {}, array2d, array2d),
("softmax_activation.v1", {}, array2d, array2d),
("Softmax.v1", {}, array2d, array2d),
("Softmax.v1", {"nO": 4, "nI": 4}, array2d, array2d),
("Softmax.v2", {}, array2d, array2d),
("Softmax.v2", {"nO": 4, "nI": 4}, array2d, array2d),
# fmt: off
# List to list
("LSTM.v1", {"bi": False}, [array2d, array2d], [array2d, array2d]),
pytest.param("PyTorchLSTM.v1", {"bi": False, "nO": width, "nI": width}, [array2d, array2d], [array2d, array2d], marks=pytest.mark.skipif(not has_torch, reason="needs PyTorch")),
# fmt: on
]
TEST_CASES = [
*TEST_CASES_SUMMABLE,
pytest.param(
"PyTorchLSTM.v1",
{"bi": True, "nO": width * 2, "nI": width},
[array2d, array2d],
[array2d, array2d],
marks=pytest.mark.skipif(not has_torch, reason="needs PyTorch"),
),
("LSTM.v1", {"bi": True}, [array2d, array2d], [array2d, array2d]),
# Ragged to array
("reduce_max.v1", {}, ragged, array2d),
("reduce_mean.v1", {}, ragged, array2d),
("reduce_sum.v1", {}, ragged, array2d),
# fmt: off
# Other
("expand_window.v1", {}, array2d, array2d),
("expand_window.v1", {}, ragged, ragged),
("Embed.v1", {"nO": 4, "nV": int(array2dint.max() + 1), "column": 0, "dropout": 0.2}, array2dint, array2d),
("Embed.v1", {"nO": 4, "nV": int(array1dint.max() + 1)}, array1dint, array2d),
("HashEmbed.v1", {"nO": 1, "nV": int(array2dint.max()), "column": 0, "dropout": 0.2}, array2dint, array2d),
("HashEmbed.v1", {"nO": 1, "nV": 2}, array1dint, array2d),
("MultiSoftmax.v1", {"nOs": (1, 3)}, array2d, array2d),
# ("CauchySimilarity.v1", {}, (array2d, array2d), array1d),
("ParametricAttention.v1", {}, ragged, ragged),
("ParametricAttention.v2", {}, ragged, ragged),
("ParametricAttention.v2", {"key_transform": {"@layers": "Gelu.v1"}}, ragged, ragged),
("SparseLinear.v1", {}, (numpy.asarray([1, 2, 3], dtype="uint64"), array1d, numpy.asarray([1, 1], dtype="i")), array2d),
("SparseLinear.v2", {}, (numpy.asarray([1, 2, 3], dtype="uint64"), array1d, numpy.asarray([1, 1], dtype="i")), array2d),
("remap_ids.v1", {"dtype": "f"}, ["a", 1, 5.0], array2dint),
("remap_ids.v2", {"mapping_table": {}, "column": 1}, numpy.array([[1, 2, 3], [4, 5, 6]]).T, array2dint),
("premap_ids.v1", {"mapping_table": {}, "column": 1}, numpy.array([[1, 2, 3], [4, 5, 6]]).T, array2dint),
# fmt: on
]
@pytest.mark.parametrize("name,kwargs,in_data,out_data", TEST_CASES)
def test_layers_from_config(name, kwargs, in_data, out_data):
cfg = {"@layers": name, **kwargs}
filled_cfg = registry.fill({"config": cfg})
assert srsly.is_json_serializable(filled_cfg)
model = registry.resolve({"config": cfg})["config"]
if "LSTM" in name:
model = with_padded(model)
valid = True
with data_validation(valid):
model.initialize(in_data, out_data)
Y, backprop = model(in_data, is_train=True)
if model.has_dim("nO"):
assert get_width(Y) == model.get_dim("nO")
assert_data_match(Y, out_data)
dX = backprop(Y)
assert_data_match(dX, in_data)
# Test that during predictions, no dropout is applied
model._to_ops(NoDropoutOps())
model.predict(in_data)
@pytest.mark.parametrize("name,kwargs,in_data,out_data", TEST_CASES_SUMMABLE)
def test_layers_with_residual(name, kwargs, in_data, out_data):
cfg = {"@layers": "residual.v1", "layer": {"@layers": name, **kwargs}}
model = registry.resolve({"config": cfg})["config"]
if "LSTM" in name:
model = with_padded(model)
model.initialize(in_data, out_data)
Y, backprop = model(in_data, is_train=True)
assert_data_match(Y, out_data)
dX = backprop(Y)
assert_data_match(dX, in_data)
@pytest.mark.parametrize("data", [array2d, ragged, padded, [array2d, array2d]])
def test_dropout(data):
model = Dropout(0.2)
model.initialize(data, data)
Y, backprop = model(data, is_train=False)
assert_data_match(Y, data)
dX = backprop(Y)
assert_data_match(dX, data)
@pytest.mark.parametrize("name,kwargs,in_data,out_data", TEST_CASES)
def test_layers_batching_all(name, kwargs, in_data, out_data):
cfg = {"@layers": name, **kwargs}
model = registry.resolve({"config": cfg})["config"]
if "expand_window" in name:
return
if "LSTM" in name:
model = with_padded(model)
util_batch_unbatch_list(model, in_data, out_data)
else:
if isinstance(in_data, OPS.xp.ndarray) and in_data.ndim == 2:
if isinstance(out_data, OPS.xp.ndarray) and out_data.ndim == 2:
util_batch_unbatch_array(model, in_data, out_data)
if isinstance(in_data, Ragged):
if isinstance(out_data, OPS.xp.ndarray) and out_data.ndim == 2:
util_batch_unbatch_ragged(model, in_data, out_data)
def util_batch_unbatch_array(
model: Model[Floats2d, Array2d], in_data: Floats2d, out_data: Array2d
):
unbatched = [model.ops.reshape2f(a, 1, -1) for a in in_data]
with data_validation(True):
model.initialize(in_data, out_data)
Y_batched = model.predict(in_data).tolist()
Y_not_batched = [model.predict(u)[0].tolist() for u in unbatched]
assert_almost_equal(Y_batched, Y_not_batched, decimal=4)
def util_batch_unbatch_list(
model: Model[List[Array2d], List[Array2d]],
in_data: List[Array2d],
out_data: List[Array2d],
):
with data_validation(True):
model.initialize(in_data, out_data)
Y_batched = model.predict(in_data)
Y_not_batched = [model.predict([u])[0] for u in in_data]
assert_almost_equal(Y_batched, Y_not_batched, decimal=4) # type: ignore
def util_batch_unbatch_ragged(
model: Model[Ragged, Array2d], in_data: Ragged, out_data: Array2d
):
with data_validation(True):
model.initialize(in_data, out_data)
Y_batched = model.predict(in_data)
Y_not_batched = [model.predict(in_data[i])[0] for i in range(len(in_data))]
assert_almost_equal(Y_batched, Y_not_batched, decimal=4) # type: ignore
|
