Skip to content

Commit 49c1e10

Browse files
kwonyulchoikwonyulchoi
committed
initial commit
1 parent 7939531 commit 49c1e10

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

42 files changed

+2094
-6
lines changed

‎.gitignore‎

Lines changed: 15 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,15 @@
1+
!jest.config.js
2+
node_modules
3+
4+
# CDK asset staging directory
5+
.cdk.staging
6+
cdk.out
7+
8+
# python
9+
**/*.pyc
10+
__pycache__
11+
.codes/**/__pycache__/
12+
13+
# vsc
14+
**/.DS_Store
15+
.vscode/**

‎README.md‎

Lines changed: 397 additions & 6 deletions
Large diffs are not rendered by default.

‎cdk.json‎

Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,17 @@
1+
{
2+
"app": "npx ts-node --prefer-ts-exts infra/app-main",
3+
"context": {
4+
"@aws-cdk/aws-apigateway:usagePlanKeyOrderInsensitiveId": true,
5+
"@aws-cdk/core:enableStackNameDuplicates": "true",
6+
"aws-cdk:enableDiffNoFail": "true",
7+
"@aws-cdk/core:stackRelativeExports": "true",
8+
"@aws-cdk/aws-ecr-assets:dockerIgnoreSupport": true,
9+
"@aws-cdk/aws-secretsmanager:parseOwnedSecretName": true,
10+
"@aws-cdk/aws-kms:defaultKeyPolicies": true,
11+
"@aws-cdk/aws-s3:grantWriteWithoutAcl": true,
12+
"@aws-cdk/aws-ecs-patterns:removeDefaultDesiredCount": true,
13+
"@aws-cdk/aws-rds:lowercaseDbIdentifier": true,
14+
"@aws-cdk/aws-efs:defaultEncryptionAtRest": true,
15+
"@aws-cdk/aws-lambda:recognizeVersionProps": true
16+
}
17+
}

‎codes/glue/churn-xgboost/script/download_data.sh‎

Lines changed: 8 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,8 @@
1+
#!/bin/sh
2+
3+
DATA_PATH=codes/glue/churn-xgboost/data
4+
DATA_FILE=input.csv
5+
6+
mkdir $DATA_PATH
7+
8+
curl -o $DATA_PATH/$DATA_FILE https://raw.githubusercontent.com/aws/amazon-sagemaker-examples/master/step-functions-data-science-sdk/automate_model_retraining_workflow/data/customer-churn.csv

‎codes/glue/churn-xgboost/script/upload_input.sh‎

Lines changed: 24 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,24 @@
1+
#!/bin/sh
2+
3+
DATA_PATH=codes/glue/churn-xgboost/data
4+
DATA_FILE=input.csv
5+
ENDPOINT_NAME=churn-xgboost
6+
7+
APP_CONFIG=$1
8+
S3_UPLOAD_KEY=$2
9+
10+
ACCOUNT=$(cat $APP_CONFIG | jq -r '.Project.Account')
11+
REGION=$(cat $APP_CONFIG | jq -r '.Project.Region')
12+
PROFILE_NAME=$(cat $APP_CONFIG | jq -r '.Project.Profile')
13+
PROJECT_NAME=$(cat $APP_CONFIG | jq -r '.Project.Name')
14+
PROJECT_STAGE=$(cat $APP_CONFIG | jq -r '.Project.Stage')
15+
PROJECT_PREFIX=$PROJECT_NAME$PROJECT_STAGE
16+
17+
PROJECT_PREFIX_LOWER=$(echo $PROJECT_PREFIX | tr '[:upper:]' '[:lower:]')
18+
ENDPOINT_NAME_LOWER=$(echo $ENDPOINT_NAME | tr '[:upper:]' '[:lower:]')
19+
20+
if [ -z "$PROFILE_NAME" ]; then
21+
aws s3 cp $DATA_PATH/$DATA_FILE s3://"$PROJECT_PREFIX_LOWER"-"$ENDPOINT_NAME_LOWER"-"$REGION"-"${ACCOUNT:0:5}"/input/"${S3_UPLOAD_KEY}"
22+
else
23+
aws s3 cp $DATA_PATH/$DATA_FILE s3://"$PROJECT_PREFIX_LOWER"-"$ENDPOINT_NAME_LOWER"-"$REGION"-"${ACCOUNT:0:5}"/input/"${S3_UPLOAD_KEY}" --profile $PROFILE_NAME
24+
fi

‎codes/glue/churn-xgboost/src/glue_etl.py‎

Lines changed: 33 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,33 @@
1+
import sys
2+
3+
from awsglue.context import GlueContext
4+
from awsglue.dynamicframe import DynamicFrame
5+
from awsglue.job import Job
6+
from awsglue.transforms import *
7+
from awsglue.utils import getResolvedOptions
8+
from pyspark.context import SparkContext
9+
10+
# Retrieve parameters for the Glue job.
11+
args = getResolvedOptions(sys.argv, ["JOB_NAME", "S3_INPUT_FILE", "S3_TRAIN_KEY", "S3_VALIDATE_KEY"])
12+
13+
sc = SparkContext()
14+
glueContext = GlueContext(sc)
15+
spark = glueContext.spark_session
16+
job = Job(glueContext)
17+
job.init(args["JOB_NAME"], args)
18+
19+
# Create a PySpark dataframe from the source table.
20+
source_data_frame = spark.read.load(args["S3_INPUT_FILE"], format="csv", inferSchema=True, header=False)
21+
22+
# Split the dataframe in to training and validation dataframes.
23+
train_data, val_data = source_data_frame.randomSplit([0.7, 0.3])
24+
25+
# Write both dataframes to the destination datastore.
26+
train_path = args["S3_TRAIN_KEY"]
27+
val_path = args["S3_VALIDATE_KEY"]
28+
29+
train_data.write.save(train_path, format="csv", mode="overwrite")
30+
val_data.write.save(val_path, format="csv", mode="overwrite")
31+
32+
# Complete the job.
33+
job.commit()

‎codes/lambda/mlops-query-accuracy/src/handler.py‎

Lines changed: 32 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,32 @@
1+
import json
2+
3+
import boto3
4+
from botocore.exceptions import ClientError
5+
6+
7+
def handle(event, context):
8+
print('handle---', event)
9+
10+
if 'TrainSageMaker' in event and 'TrainingJobName' in event['TrainSageMaker']:
11+
client = boto3.client('sagemaker')
12+
training_job_name = event['TrainSageMaker']['TrainingJobName']
13+
14+
try:
15+
response = client.describe_training_job(
16+
TrainingJobName=training_job_name
17+
)
18+
print('response', response)
19+
for metric in response["FinalMetricDataList"]:
20+
metric["Timestamp"] = metric["Timestamp"].timestamp()
21+
22+
print('Metric ==> {}'.format(json.dumps(response["FinalMetricDataList"])))
23+
return {'Metrics': response["FinalMetricDataList"]}
24+
except ClientError as e:
25+
error_msg = '{} is not existent in training list'.format(training_job_name)
26+
print(error_msg)
27+
raise KeyError(error_msg)
28+
else:
29+
error_msg = 'key error: check key in event - {}'.format(json.dumps(event))
30+
print(error_msg)
31+
raise KeyError(error_msg)
32+

‎codes/lambda/mlops-query-endpoint/src/handler.py‎

Lines changed: 27 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,27 @@
1+
import json
2+
3+
import boto3
4+
from botocore.exceptions import ClientError
5+
6+
7+
def handle(event, context):
8+
print('handle---', event)
9+
10+
if 'ServeSageMaker' in event and 'EndpointName' in event['ServeSageMaker']:
11+
client = boto3.client('sagemaker')
12+
endpoint_name = event['ServeSageMaker']['EndpointName']
13+
14+
try:
15+
response = client.describe_endpoint(
16+
EndpointName=endpoint_name
17+
)
18+
print('response', response)
19+
return {'Existent': 'TRUE'}
20+
except ClientError as e:
21+
print('{} is not existent in endpoint list'.format(endpoint_name))
22+
return {'Existent': 'FALSE'}
23+
else:
24+
error_msg = 'key error: check key in event - {}'.format(json.dumps(event))
25+
print(error_msg)
26+
raise KeyError(error_msg)
27+

‎codes/lambda/mlops-trigger-statemachine/src/handler.py‎

Lines changed: 82 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,82 @@
1+
import os
2+
import sys
3+
import json
4+
import time
5+
6+
import boto3
7+
from botocore.exceptions import ClientError
8+
9+
10+
_endpoint_name = os.environ.get('EndpointName', 'no-endpoint-name')
11+
_statemachine_arn = os.environ.get('StateMachineArn', 'no-statemachine-arn')
12+
_bucket_key_for_input = os.environ.get('BucketKeyForInput', 'no-bucket-key-for-input')
13+
_bucket_key_for_output = os.environ.get('BucketKeyForOutput', 'no-bucket-key-for-output')
14+
15+
16+
def trigger_statemachine(client, execution_name, statemachine_arn, request):
17+
print('trigger_statemachine: request==>', request)
18+
19+
if len(execution_name) > 80:
20+
execution_name = execution_name[:80]
21+
22+
try:
23+
client.start_execution(stateMachineArn=statemachine_arn,
24+
name=execution_name, input=json.dumps(request))
25+
except ClientError as e:
26+
print('Error : trigger_statemachine - {}'.format(e.response['Error']['Message']))
27+
28+
29+
def get_request_template():
30+
return {
31+
"PreprocessGlue": {
32+
"--S3_INPUT_FILE": "s3://bucket-name/input/xxx/input.csv",
33+
"--S3_TRAIN_KEY": "s3://bucket-name/output/xxx/train/",
34+
"--S3_VALIDATE_KEY": "s3://bucket-name/output/xxx/validate/"
35+
},
36+
"TrainSageMaker": {
37+
"TrainingJobName": "projectprefix-xxx-input",
38+
"TrainData": "s3://bucket-name/output/xxx/train/",
39+
"ValidateData": "s3://bucket-name/output/xxx/validate/",
40+
"TrainOutput": "s3://bucket-name/output/xxx/model/projectprefix-xxx-input/"
41+
},
42+
"ServeSageMaker": {
43+
"ModelName": "projectprefix-xxx-input",
44+
"EndpointConfigName": "projectprefix-xxx-input",
45+
"EndpointName": "churn-xgboost"
46+
}
47+
}
48+
49+
50+
def handle(event, context):
51+
print('event--->', event)
52+
53+
client = boto3.client('stepfunctions')
54+
55+
for record in event['Records']:
56+
bucket_name: str = record['s3']['bucket']['name']
57+
input_file_key: str = record['s3']['object']['key']
58+
print('etl_input_file_key => {}'.format(input_file_key))
59+
60+
if input_file_key.find('.') > -1:
61+
temp = input_file_key.replace(f'{_bucket_key_for_input}/', f'{_bucket_key_for_output}/', 1)
62+
output_dir_key = temp.split('.')[0]
63+
64+
temp = output_dir_key.replace(f'{_bucket_key_for_output}/', '', 1)
65+
temp = temp.replace("/", "-")
66+
execution_name = f'{_endpoint_name}-{temp}'
67+
68+
request = get_request_template()
69+
request['PreprocessGlue']['--S3_INPUT_FILE'] = f's3://{bucket_name}/{input_file_key}'
70+
request['PreprocessGlue']['--S3_TRAIN_KEY'] = f's3://{bucket_name}/{output_dir_key}/data/train/'
71+
request['PreprocessGlue']['--S3_VALIDATE_KEY'] = f's3://{bucket_name}/{output_dir_key}/data/validate/'
72+
73+
request['TrainSageMaker']['TrainingJobName'] = execution_name
74+
request['TrainSageMaker']['TrainData'] = request['PreprocessGlue']['--S3_TRAIN_KEY']
75+
request['TrainSageMaker']['ValidateData'] = request['PreprocessGlue']['--S3_VALIDATE_KEY']
76+
request['TrainSageMaker']['TrainOutput'] = f's3://{bucket_name}/{output_dir_key}/model/'
77+
78+
request['ServeSageMaker']['ModelName'] = execution_name
79+
request['ServeSageMaker']['EndpointConfigName'] = execution_name
80+
request['ServeSageMaker']['EndpointName'] = _endpoint_name
81+
82+
trigger_statemachine(client, execution_name, _statemachine_arn, request)

‎config/app-config-demo.json‎

Lines changed: 40 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,40 @@
1+
{
2+
"Project": {
3+
"Name": "MLOpsPipeline",
4+
"Stage": "Demo",
5+
"Account": "75157*******",
6+
"Region": "us-east-2",
7+
"Profile": "cdk-demo"
8+
},
9+
10+
"Stack": {
11+
"ChurnXgboostPipeline": {
12+
"Name": "ChurnXgboostPipelineStack",
13+
14+
"EndpointName": "churn-xgboost",
15+
16+
"GlueJobFilePath": "codes/glue/churn-xgboost/src/glue_etl.py",
17+
"GlueJobTimeoutInMin": 30,
18+
19+
"TrainContainerImage": "825641698319.dkr.ecr.us-east-2.amazonaws.com/xgboost:1",
20+
"TrainParameters": {
21+
"max_depth": "5",
22+
"eta": "0.2",
23+
"gamma": "4",
24+
"min_child_weight": "6",
25+
"subsample": "0.7",
26+
"objective": "multi:softprob",
27+
"num_class": "2",
28+
"num_round": "50"
29+
},
30+
"TrainInputContent": "text/csv",
31+
"TrainInstanceType": "c5.xlarge",
32+
33+
"ModelValidationEnable": true,
34+
"ModelErrorThreshold": 0.1,
35+
36+
"EndpointInstanceType": "t2.2xlarge",
37+
"EndpointInstanceCount": 1
38+
}
39+
}
40+
}

0 commit comments

Comments
 (0)