/
solutionsteam_tf_release_supported.py
180 lines (163 loc) · 5.85 KB
/
solutionsteam_tf_release_supported.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
# Copyright 2024 Google LLC
#
# 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.
"""A DAG to run all supported ML models with the nightly TensorFlow version."""
import datetime
from airflow import models
from dags import composer_env
from dags.vm_resource import TpuVersion, Project, Zone, RuntimeVersion, V5_NETWORKS, V5E_SUBNETWORKS, V5P_SUBNETWORKS
from dags.solutions_team.configs.tensorflow import solutionsteam_tf_release_supported_config as tf_config
from dags.solutions_team.configs.tensorflow import common
# Release tests only need to run once, they can be run manually as needed
SCHEDULED_TIME = None
VERSION = f"{tf_config.MAJOR_VERSION}.{tf_config.MINOR_VERSION}"
with models.DAG(
dag_id=f"tf_{tf_config.MAJOR_VERSION}_{tf_config.MINOR_VERSION}_nightly_supported",
schedule=SCHEDULED_TIME,
tags=["solutions_team", "tf", VERSION, "supported", "xlml"],
start_date=datetime.datetime(2023, 8, 16),
catchup=False,
) as dag:
# Keras - tests run in sequence order
tf_keras_v2_8 = []
for feature, name in common.FEATURE_NAME.items():
test = tf_config.get_tf_keras_config(
tpu_version=TpuVersion.V2,
tpu_cores=8,
tpu_zone=Zone.US_CENTRAL1_C.value,
time_out_in_min=common.FEATURE_TIMEOUT.get(feature),
test_feature=feature,
test_name=name,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5.value,
).run()
if tf_keras_v2_8:
tf_keras_v2_8[-1] >> test
tf_keras_v2_8.append(test)
tf_keras_v5e_4 = []
for feature, name in common.FEATURE_NAME.items():
test = tf_config.get_tf_keras_config(
project_name=Project.TPU_PROD_ENV_AUTOMATED.value,
tpu_version=TpuVersion.V5E,
tpu_cores=4,
tpu_zone=Zone.US_EAST1_C.value,
time_out_in_min=common.FEATURE_TIMEOUT.get(feature),
test_feature=feature,
test_name=name,
network=V5_NETWORKS,
subnetwork=V5E_SUBNETWORKS,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5_LITE.value,
).run()
if tf_keras_v5e_4:
tf_keras_v5e_4[-1] >> test
tf_keras_v5e_4.append(test)
tf_keras_v5p_8 = []
for feature, name in common.FEATURE_NAME.items():
test = tf_config.get_tf_keras_config(
project_name=Project.TPU_PROD_ENV_AUTOMATED.value,
tpu_version=TpuVersion.V5P,
tpu_cores=8,
tpu_zone=Zone.US_EAST5_A.value,
time_out_in_min=common.FEATURE_TIMEOUT.get(feature),
test_feature=feature,
test_name=name,
network=V5_NETWORKS,
subnetwork=V5P_SUBNETWORKS,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5.value,
).run()
if tf_keras_v5p_8:
tf_keras_v5p_8[-1] >> test
tf_keras_v5p_8.append(test)
# ResNet
tf_resnet_v2_8 = tf_config.get_tf_resnet_config(
tpu_version=TpuVersion.V2,
tpu_cores=8,
tpu_zone=Zone.US_CENTRAL1_C.value,
time_out_in_min=60,
global_batch_size=1024,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5.value,
).run()
tf_resnet_v3_8 = tf_config.get_tf_resnet_config(
tpu_version=TpuVersion.V3,
tpu_cores=8,
tpu_zone=Zone.US_EAST1_D.value,
time_out_in_min=60,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5.value,
).run()
tf_resnet_v4_8 = tf_config.get_tf_resnet_config(
tpu_version=TpuVersion.V4,
tpu_cores=8,
tpu_zone=Zone.US_CENTRAL2_B.value,
time_out_in_min=60,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5.value,
).run()
tf_resnet_v4_32 = tf_config.get_tf_resnet_config(
tpu_version=TpuVersion.V4,
tpu_cores=32,
tpu_zone=Zone.US_CENTRAL2_B.value,
time_out_in_min=60,
is_pod=True,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5.value,
).run()
tf_resnet_v5e_4 = tf_config.get_tf_resnet_config(
project_name=Project.TPU_PROD_ENV_AUTOMATED.value,
tpu_version=TpuVersion.V5E,
tpu_cores=4,
tpu_zone=Zone.US_EAST1_C.value,
time_out_in_min=60,
global_batch_size=2048,
network=V5_NETWORKS,
subnetwork=V5E_SUBNETWORKS,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5_LITE.value,
).run()
tf_resnet_v5e_16 = tf_config.get_tf_resnet_config(
project_name=Project.TPU_PROD_ENV_AUTOMATED.value,
tpu_version=TpuVersion.V5E,
tpu_cores=16,
tpu_zone=Zone.US_EAST1_C.value,
time_out_in_min=60,
global_batch_size=2048,
network=V5_NETWORKS,
subnetwork=V5E_SUBNETWORKS,
is_pod=True,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5_LITE.value,
).run()
tf_resnet_v5p_8 = tf_config.get_tf_resnet_config(
project_name=Project.TPU_PROD_ENV_AUTOMATED.value,
tpu_version=TpuVersion.V5P,
tpu_cores=8,
tpu_zone=Zone.US_EAST5_A.value,
time_out_in_min=60,
network=V5_NETWORKS,
subnetwork=V5P_SUBNETWORKS,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5.value,
).run()
tf_resnet_v5p_32 = tf_config.get_tf_resnet_config(
project_name=Project.TPU_PROD_ENV_AUTOMATED.value,
tpu_version=TpuVersion.V5P,
tpu_cores=32,
tpu_zone=Zone.US_EAST5_A.value,
time_out_in_min=60,
network=V5_NETWORKS,
subnetwork=V5P_SUBNETWORKS,
is_pod=True,
runtime_version=RuntimeVersion.V2_ALPHA_TPUV5.value,
).run()
# Test dependencies
tf_keras_v2_8
tf_keras_v5e_4
tf_keras_v5p_8
tf_resnet_v2_8
tf_resnet_v3_8
tf_resnet_v4_8 >> tf_resnet_v4_32
tf_resnet_v5e_4 >> tf_resnet_v5e_16
tf_resnet_v5p_8 >> tf_resnet_v5p_32