forked from huggingface/transformers
-
Notifications
You must be signed in to change notification settings - Fork 0
/
test_benchmark_tf.py
226 lines (206 loc) · 8.84 KB
/
test_benchmark_tf.py
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
# Copyright 2020 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import tempfile
import unittest
from pathlib import Path
from transformers import AutoConfig, is_tf_available
from transformers.testing_utils import require_tf
if is_tf_available():
import tensorflow as tf
from transformers import TensorFlowBenchmark, TensorFlowBenchmarkArguments
@require_tf
class TFBenchmarkTest(unittest.TestCase):
def check_results_dict_not_empty(self, results):
for model_result in results.values():
for batch_size, sequence_length in zip(model_result["bs"], model_result["ss"]):
result = model_result["result"][batch_size][sequence_length]
self.assertIsNotNone(result)
def test_inference_no_configs_eager(self):
MODEL_ID = "sshleifer/tiny-gpt2"
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
training=False,
inference=True,
sequence_lengths=[8],
batch_sizes=[1],
eager_mode=True,
multi_process=False,
)
benchmark = TensorFlowBenchmark(benchmark_args)
results = benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result)
self.check_results_dict_not_empty(results.memory_inference_result)
def test_inference_no_configs_only_pretrain(self):
MODEL_ID = "sshleifer/tiny-distilbert-base-uncased-finetuned-sst-2-english"
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
training=False,
inference=True,
sequence_lengths=[8],
batch_sizes=[1],
multi_process=False,
only_pretrain_model=True,
)
benchmark = TensorFlowBenchmark(benchmark_args)
results = benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result)
self.check_results_dict_not_empty(results.memory_inference_result)
def test_inference_no_configs_graph(self):
MODEL_ID = "sshleifer/tiny-gpt2"
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
training=False,
inference=True,
sequence_lengths=[8],
batch_sizes=[1],
multi_process=False,
)
benchmark = TensorFlowBenchmark(benchmark_args)
results = benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result)
self.check_results_dict_not_empty(results.memory_inference_result)
def test_inference_with_configs_eager(self):
MODEL_ID = "sshleifer/tiny-gpt2"
config = AutoConfig.from_pretrained(MODEL_ID)
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
training=False,
inference=True,
sequence_lengths=[8],
batch_sizes=[1],
eager_mode=True,
multi_process=False,
)
benchmark = TensorFlowBenchmark(benchmark_args, [config])
results = benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result)
self.check_results_dict_not_empty(results.memory_inference_result)
def test_inference_with_configs_graph(self):
MODEL_ID = "sshleifer/tiny-gpt2"
config = AutoConfig.from_pretrained(MODEL_ID)
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
training=False,
inference=True,
sequence_lengths=[8],
batch_sizes=[1],
multi_process=False,
)
benchmark = TensorFlowBenchmark(benchmark_args, [config])
results = benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result)
self.check_results_dict_not_empty(results.memory_inference_result)
def test_train_no_configs(self):
MODEL_ID = "sshleifer/tiny-gpt2"
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
training=True,
inference=False,
sequence_lengths=[8],
batch_sizes=[1],
multi_process=False,
)
benchmark = TensorFlowBenchmark(benchmark_args)
results = benchmark.run()
self.check_results_dict_not_empty(results.time_train_result)
self.check_results_dict_not_empty(results.memory_train_result)
def test_train_with_configs(self):
MODEL_ID = "sshleifer/tiny-gpt2"
config = AutoConfig.from_pretrained(MODEL_ID)
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
training=True,
inference=False,
sequence_lengths=[8],
batch_sizes=[1],
multi_process=False,
)
benchmark = TensorFlowBenchmark(benchmark_args, [config])
results = benchmark.run()
self.check_results_dict_not_empty(results.time_train_result)
self.check_results_dict_not_empty(results.memory_train_result)
def test_inference_encoder_decoder_with_configs(self):
MODEL_ID = "patrickvonplaten/t5-tiny-random"
config = AutoConfig.from_pretrained(MODEL_ID)
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
training=False,
inference=True,
sequence_lengths=[8],
batch_sizes=[1],
multi_process=False,
)
benchmark = TensorFlowBenchmark(benchmark_args, configs=[config])
results = benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result)
self.check_results_dict_not_empty(results.memory_inference_result)
@unittest.skipIf(is_tf_available() and len(tf.config.list_physical_devices("GPU")) == 0, "Cannot do xla on CPU.")
def test_inference_no_configs_xla(self):
MODEL_ID = "sshleifer/tiny-gpt2"
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
training=False,
inference=True,
sequence_lengths=[8],
batch_sizes=[1],
use_xla=True,
multi_process=False,
)
benchmark = TensorFlowBenchmark(benchmark_args)
results = benchmark.run()
self.check_results_dict_not_empty(results.time_inference_result)
self.check_results_dict_not_empty(results.memory_inference_result)
def test_save_csv_files(self):
MODEL_ID = "sshleifer/tiny-gpt2"
with tempfile.TemporaryDirectory() as tmp_dir:
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
inference=True,
save_to_csv=True,
sequence_lengths=[8],
batch_sizes=[1],
inference_time_csv_file=os.path.join(tmp_dir, "inf_time.csv"),
inference_memory_csv_file=os.path.join(tmp_dir, "inf_mem.csv"),
env_info_csv_file=os.path.join(tmp_dir, "env.csv"),
multi_process=False,
)
benchmark = TensorFlowBenchmark(benchmark_args)
benchmark.run()
self.assertTrue(Path(os.path.join(tmp_dir, "inf_time.csv")).exists())
self.assertTrue(Path(os.path.join(tmp_dir, "inf_mem.csv")).exists())
self.assertTrue(Path(os.path.join(tmp_dir, "env.csv")).exists())
def test_trace_memory(self):
MODEL_ID = "sshleifer/tiny-gpt2"
def _check_summary_is_not_empty(summary):
self.assertTrue(hasattr(summary, "sequential"))
self.assertTrue(hasattr(summary, "cumulative"))
self.assertTrue(hasattr(summary, "current"))
self.assertTrue(hasattr(summary, "total"))
with tempfile.TemporaryDirectory() as tmp_dir:
benchmark_args = TensorFlowBenchmarkArguments(
models=[MODEL_ID],
inference=True,
sequence_lengths=[8],
batch_sizes=[1],
log_filename=os.path.join(tmp_dir, "log.txt"),
log_print=True,
trace_memory_line_by_line=True,
eager_mode=True,
multi_process=False,
)
benchmark = TensorFlowBenchmark(benchmark_args)
result = benchmark.run()
_check_summary_is_not_empty(result.inference_summary)
self.assertTrue(Path(os.path.join(tmp_dir, "log.txt")).exists())