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
|
from sagemaker.workflow.pipeline import Pipeline
from sagemaker.workflow.steps import TrainingStep, ProcessingStep
from sagemaker.workflow.parameters import ParameterString, ParameterFloat
class MLOpsPipeline:
def __init__(self, session, role):
self.session = session
self.role = role
# パイプライン パラメータ
self.model_name = ParameterString(name="ModelName", default_value="customer-churn-model")
self.train_instance_type = ParameterString(name="TrainInstanceType", default_value="ml.m5.xlarge")
self.learning_rate = ParameterFloat(name="LearningRate", default_value=0.001)
def create_preprocessing_step(self):
"""データ前処理ステップ"""
from sagemaker.sklearn.processing import SKLearnProcessor
processor = SKLearnProcessor(
framework_version='1.0-1',
instance_type='ml.m5.large',
instance_count=1,
base_job_name='preprocessing',
role=self.role
)
processing_step = ProcessingStep(
name="PreprocessingStep",
processor=processor,
code='preprocessing.py',
inputs=[
sagemaker.processing.ProcessingInput(
source='s3://my-bucket/raw-data/',
destination='/opt/ml/processing/input'
)
],
outputs=[
sagemaker.processing.ProcessingOutput(
output_name='train_data',
source='/opt/ml/processing/train',
destination='s3://my-bucket/processed/train'
),
sagemaker.processing.ProcessingOutput(
output_name='validation_data',
source='/opt/ml/processing/validation',
destination='s3://my-bucket/processed/validation'
)
]
)
return processing_step
def create_training_step(self, preprocessing_step):
"""モデル訓練ステップ"""
estimator = PyTorch(
entry_point='train.py',
source_dir='src',
role=self.role,
instance_type=self.train_instance_type,
framework_version='2.0.0',
py_version='py310',
hyperparameters={
'learning_rate': self.learning_rate,
'epochs': 50,
'batch_size': 32
}
)
training_step = TrainingStep(
name="TrainingStep",
estimator=estimator,
inputs={
'train': TrainingInput(
s3_data=preprocessing_step.properties.ProcessingOutputConfig.Outputs['train_data'].S3Output.S3Uri
),
'validation': TrainingInput(
s3_data=preprocessing_step.properties.ProcessingOutputConfig.Outputs['validation_data'].S3Output.S3Uri
)
}
)
return training_step
def create_evaluation_step(self, training_step):
"""モデル評価ステップ"""
from sagemaker.workflow.steps import ProcessingStep
from sagemaker.workflow.conditions import ConditionGreaterThanOrEqualTo
from sagemaker.workflow.condition_step import ConditionStep
from sagemaker.workflow.properties import PropertyFile
# 評価処理
evaluation_processor = SKLearnProcessor(
framework_version='1.0-1',
instance_type='ml.m5.large',
instance_count=1,
role=self.role
)
evaluation_step = ProcessingStep(
name="EvaluationStep",
processor=evaluation_processor,
code='evaluate.py',
inputs=[
ProcessingInput(
source=training_step.properties.ModelArtifacts.S3ModelArtifacts,
destination='/opt/ml/processing/model'
)
],
outputs=[
ProcessingOutput(
output_name='evaluation_results',
source='/opt/ml/processing/evaluation',
destination='s3://my-bucket/evaluation'
)
],
property_files=[
PropertyFile(
name="EvaluationReport",
output_name="evaluation_results",
path="evaluation.json"
)
]
)
return evaluation_step
def build_pipeline(self):
"""パイプライン構築"""
# ステップ定義
preprocessing_step = self.create_preprocessing_step()
training_step = self.create_training_step(preprocessing_step)
evaluation_step = self.create_evaluation_step(training_step)
# 条件分岐（モデル品質チェック）
model_quality_condition = ConditionGreaterThanOrEqualTo(
left=JsonGet(
step_name=evaluation_step.name,
property_file="EvaluationReport",
json_path="metrics.accuracy"
),
right=0.8 # 精度80%以上
)
# 条件に基づくモデル登録
condition_step = ConditionStep(
name="ModelQualityCheck",
conditions=[model_quality_condition],
if_steps=[self.create_model_registration_step(training_step)],
else_steps=[]
)
# パイプライン作成
pipeline = Pipeline(
name=f"MLPipeline-{self.model_name.default_value}",
parameters=[
self.model_name,
self.train_instance_type,
self.learning_rate
],
steps=[
preprocessing_step,
training_step,
evaluation_step,
condition_step
]
)
return pipeline
|
