408 lines
14 KiB
Python
408 lines
14 KiB
Python
import datetime
|
|
import re
|
|
|
|
import boto3
|
|
from botocore.exceptions import ClientError
|
|
import pytest
|
|
|
|
from moto import mock_sagemaker
|
|
from moto.core import DEFAULT_ACCOUNT_ID as ACCOUNT_ID
|
|
|
|
FAKE_ROLE_ARN = f"arn:aws:iam::{ACCOUNT_ID}:role/FakeRole"
|
|
FAKE_PROCESSING_JOB_NAME = "MyProcessingJob"
|
|
FAKE_CONTAINER = "382416733822.dkr.ecr.us-east-1.amazonaws.com/linear-learner:1"
|
|
TEST_REGION_NAME = "us-east-1"
|
|
|
|
|
|
@pytest.fixture(name="sagemaker_client")
|
|
def fixture_sagemaker_client():
|
|
with mock_sagemaker():
|
|
yield boto3.client("sagemaker", region_name=TEST_REGION_NAME)
|
|
|
|
|
|
class MyProcessingJobModel:
|
|
def __init__(
|
|
self,
|
|
processing_job_name,
|
|
role_arn,
|
|
container=None,
|
|
bucket=None,
|
|
prefix=None,
|
|
app_specification=None,
|
|
network_config=None,
|
|
processing_inputs=None,
|
|
processing_output_config=None,
|
|
processing_resources=None,
|
|
stopping_condition=None,
|
|
):
|
|
self.processing_job_name = processing_job_name
|
|
self.role_arn = role_arn
|
|
self.container = (
|
|
container
|
|
or "683313688378.dkr.ecr.us-east-1.amazonaws.com/sagemaker-scikit-learn:0.23-1-cpu-py3"
|
|
)
|
|
self.bucket = bucket or "my-bucket"
|
|
self.prefix = prefix or "sagemaker"
|
|
self.app_specification = app_specification or {
|
|
"ImageUri": self.container,
|
|
"ContainerEntrypoint": ["python3"],
|
|
}
|
|
self.network_config = network_config or {
|
|
"EnableInterContainerTrafficEncryption": False,
|
|
"EnableNetworkIsolation": False,
|
|
}
|
|
self.processing_inputs = processing_inputs or [
|
|
{
|
|
"InputName": "input",
|
|
"AppManaged": False,
|
|
"S3Input": {
|
|
"S3Uri": f"s3://{self.bucket}/{self.prefix}/processing/",
|
|
"LocalPath": "/opt/ml/processing/input",
|
|
"S3DataType": "S3Prefix",
|
|
"S3InputMode": "File",
|
|
"S3DataDistributionType": "FullyReplicated",
|
|
"S3CompressionType": "None",
|
|
},
|
|
}
|
|
]
|
|
self.processing_output_config = processing_output_config or {
|
|
"Outputs": [
|
|
{
|
|
"OutputName": "output",
|
|
"S3Output": {
|
|
"S3Uri": f"s3://{self.bucket}/{self.prefix}/processing/",
|
|
"LocalPath": "/opt/ml/processing/output",
|
|
"S3UploadMode": "EndOfJob",
|
|
},
|
|
"AppManaged": False,
|
|
}
|
|
]
|
|
}
|
|
self.processing_resources = processing_resources or {
|
|
"ClusterConfig": {
|
|
"InstanceCount": 1,
|
|
"InstanceType": "ml.m5.large",
|
|
"VolumeSizeInGB": 10,
|
|
},
|
|
}
|
|
self.stopping_condition = stopping_condition or {
|
|
"MaxRuntimeInSeconds": 3600,
|
|
}
|
|
|
|
def save(self, sagemaker_client):
|
|
params = {
|
|
"AppSpecification": self.app_specification,
|
|
"NetworkConfig": self.network_config,
|
|
"ProcessingInputs": self.processing_inputs,
|
|
"ProcessingJobName": self.processing_job_name,
|
|
"ProcessingOutputConfig": self.processing_output_config,
|
|
"ProcessingResources": self.processing_resources,
|
|
"RoleArn": self.role_arn,
|
|
"StoppingCondition": self.stopping_condition,
|
|
}
|
|
|
|
return sagemaker_client.create_processing_job(**params)
|
|
|
|
|
|
def test_create_processing_job(sagemaker_client):
|
|
bucket = "my-bucket"
|
|
prefix = "my-prefix"
|
|
app_specification = {
|
|
"ImageUri": FAKE_CONTAINER,
|
|
"ContainerEntrypoint": ["python3", "app.py"],
|
|
}
|
|
processing_resources = {
|
|
"ClusterConfig": {
|
|
"InstanceCount": 2,
|
|
"InstanceType": "ml.m5.xlarge",
|
|
"VolumeSizeInGB": 20,
|
|
},
|
|
}
|
|
stopping_condition = {"MaxRuntimeInSeconds": 60 * 60}
|
|
|
|
job = MyProcessingJobModel(
|
|
processing_job_name=FAKE_PROCESSING_JOB_NAME,
|
|
role_arn=FAKE_ROLE_ARN,
|
|
container=FAKE_CONTAINER,
|
|
bucket=bucket,
|
|
prefix=prefix,
|
|
app_specification=app_specification,
|
|
processing_resources=processing_resources,
|
|
stopping_condition=stopping_condition,
|
|
)
|
|
resp = job.save(sagemaker_client)
|
|
assert re.match(
|
|
rf"^arn:aws:sagemaker:.*:.*:processing-job/{FAKE_PROCESSING_JOB_NAME}$",
|
|
resp["ProcessingJobArn"],
|
|
)
|
|
|
|
resp = sagemaker_client.describe_processing_job(
|
|
ProcessingJobName=FAKE_PROCESSING_JOB_NAME
|
|
)
|
|
assert resp["ProcessingJobName"] == FAKE_PROCESSING_JOB_NAME
|
|
assert re.match(
|
|
rf"^arn:aws:sagemaker:.*:.*:processing-job/{FAKE_PROCESSING_JOB_NAME}$",
|
|
resp["ProcessingJobArn"],
|
|
)
|
|
assert "python3" in resp["AppSpecification"]["ContainerEntrypoint"]
|
|
assert "app.py" in resp["AppSpecification"]["ContainerEntrypoint"]
|
|
assert resp["RoleArn"] == FAKE_ROLE_ARN
|
|
assert resp["ProcessingJobStatus"] == "Completed"
|
|
assert isinstance(resp["CreationTime"], datetime.datetime)
|
|
assert isinstance(resp["LastModifiedTime"], datetime.datetime)
|
|
|
|
|
|
def test_list_processing_jobs(sagemaker_client):
|
|
test_processing_job = MyProcessingJobModel(
|
|
processing_job_name=FAKE_PROCESSING_JOB_NAME, role_arn=FAKE_ROLE_ARN
|
|
)
|
|
test_processing_job.save(sagemaker_client)
|
|
processing_jobs = sagemaker_client.list_processing_jobs()
|
|
assert len(processing_jobs["ProcessingJobSummaries"]) == 1
|
|
assert (
|
|
processing_jobs["ProcessingJobSummaries"][0]["ProcessingJobName"]
|
|
== FAKE_PROCESSING_JOB_NAME
|
|
)
|
|
|
|
assert re.match(
|
|
rf"^arn:aws:sagemaker:.*:.*:processing-job/{FAKE_PROCESSING_JOB_NAME}$",
|
|
processing_jobs["ProcessingJobSummaries"][0]["ProcessingJobArn"],
|
|
)
|
|
assert processing_jobs.get("NextToken") is None
|
|
|
|
|
|
def test_list_processing_jobs_multiple(sagemaker_client):
|
|
name_job_1 = "blah"
|
|
arn_job_1 = "arn:aws:sagemaker:us-east-1:000000000000:x-x/foobar"
|
|
test_processing_job_1 = MyProcessingJobModel(
|
|
processing_job_name=name_job_1, role_arn=arn_job_1
|
|
)
|
|
test_processing_job_1.save(sagemaker_client)
|
|
|
|
name_job_2 = "blah2"
|
|
arn_job_2 = "arn:aws:sagemaker:us-east-1:000000000000:x-x/foobar2"
|
|
test_processing_job_2 = MyProcessingJobModel(
|
|
processing_job_name=name_job_2, role_arn=arn_job_2
|
|
)
|
|
test_processing_job_2.save(sagemaker_client)
|
|
processing_jobs_limit = sagemaker_client.list_processing_jobs(MaxResults=1)
|
|
assert len(processing_jobs_limit["ProcessingJobSummaries"]) == 1
|
|
|
|
processing_jobs = sagemaker_client.list_processing_jobs()
|
|
assert len(processing_jobs["ProcessingJobSummaries"]) == 2
|
|
assert processing_jobs.get("NextToken") is None
|
|
|
|
|
|
def test_list_processing_jobs_none(sagemaker_client):
|
|
processing_jobs = sagemaker_client.list_processing_jobs()
|
|
assert len(processing_jobs["ProcessingJobSummaries"]) == 0
|
|
|
|
|
|
def test_list_processing_jobs_should_validate_input(sagemaker_client):
|
|
junk_status_equals = "blah"
|
|
with pytest.raises(ClientError) as ex:
|
|
sagemaker_client.list_processing_jobs(StatusEquals=junk_status_equals)
|
|
expected_error = (
|
|
f"1 validation errors detected: Value '{junk_status_equals}' at "
|
|
"'statusEquals' failed to satisfy constraint: Member must satisfy "
|
|
"enum value set: ['Completed', 'Stopped', 'InProgress', 'Stopping', "
|
|
"'Failed']"
|
|
)
|
|
assert ex.value.response["Error"]["Code"] == "ValidationException"
|
|
assert ex.value.response["Error"]["Message"] == expected_error
|
|
|
|
junk_next_token = "asdf"
|
|
with pytest.raises(ClientError) as ex:
|
|
sagemaker_client.list_processing_jobs(NextToken=junk_next_token)
|
|
assert ex.value.response["Error"]["Code"] == "ValidationException"
|
|
assert (
|
|
ex.value.response["Error"]["Message"]
|
|
== 'Invalid pagination token because "{0}".'
|
|
)
|
|
|
|
|
|
def test_list_processing_jobs_with_name_filters(sagemaker_client):
|
|
for i in range(5):
|
|
name = f"xgboost-{i}"
|
|
arn = f"arn:aws:sagemaker:us-east-1:000000000000:x-x/foobar-{i}"
|
|
MyProcessingJobModel(processing_job_name=name, role_arn=arn).save(
|
|
sagemaker_client
|
|
)
|
|
|
|
for i in range(5):
|
|
name = f"vgg-{i}"
|
|
arn = f"arn:aws:sagemaker:us-east-1:000000000000:x-x/barfoo-{i}"
|
|
MyProcessingJobModel(processing_job_name=name, role_arn=arn).save(
|
|
sagemaker_client
|
|
)
|
|
|
|
xgboost_processing_jobs = sagemaker_client.list_processing_jobs(
|
|
NameContains="xgboost"
|
|
)
|
|
assert len(xgboost_processing_jobs["ProcessingJobSummaries"]) == 5
|
|
|
|
processing_jobs_with_2 = sagemaker_client.list_processing_jobs(NameContains="2")
|
|
assert len(processing_jobs_with_2["ProcessingJobSummaries"]) == 2
|
|
|
|
|
|
def test_list_processing_jobs_paginated(sagemaker_client):
|
|
for i in range(5):
|
|
name = f"xgboost-{i}"
|
|
arn = f"arn:aws:sagemaker:us-east-1:000000000000:x-x/foobar-{i}"
|
|
MyProcessingJobModel(processing_job_name=name, role_arn=arn).save(
|
|
sagemaker_client
|
|
)
|
|
|
|
xgboost_processing_job_1 = sagemaker_client.list_processing_jobs(
|
|
NameContains="xgboost", MaxResults=1
|
|
)
|
|
assert len(xgboost_processing_job_1["ProcessingJobSummaries"]) == 1
|
|
assert (
|
|
xgboost_processing_job_1["ProcessingJobSummaries"][0]["ProcessingJobName"]
|
|
== "xgboost-0"
|
|
)
|
|
assert xgboost_processing_job_1.get("NextToken") is not None
|
|
|
|
xgboost_processing_job_next = sagemaker_client.list_processing_jobs(
|
|
NameContains="xgboost",
|
|
MaxResults=1,
|
|
NextToken=xgboost_processing_job_1.get("NextToken"),
|
|
)
|
|
assert len(xgboost_processing_job_next["ProcessingJobSummaries"]) == 1
|
|
assert (
|
|
xgboost_processing_job_next["ProcessingJobSummaries"][0]["ProcessingJobName"]
|
|
== "xgboost-1"
|
|
)
|
|
assert xgboost_processing_job_next.get("NextToken") is not None
|
|
|
|
|
|
def test_list_processing_jobs_paginated_with_target_in_middle(sagemaker_client):
|
|
for i in range(5):
|
|
name = f"xgboost-{i}"
|
|
arn = f"arn:aws:sagemaker:us-east-1:000000000000:x-x/foobar-{i}"
|
|
MyProcessingJobModel(processing_job_name=name, role_arn=arn).save(
|
|
sagemaker_client
|
|
)
|
|
|
|
for i in range(5):
|
|
name = f"vgg-{i}"
|
|
arn = f"arn:aws:sagemaker:us-east-1:000000000000:x-x/barfoo-{i}"
|
|
MyProcessingJobModel(processing_job_name=name, role_arn=arn).save(
|
|
sagemaker_client
|
|
)
|
|
|
|
vgg_processing_job_1 = sagemaker_client.list_processing_jobs(
|
|
NameContains="vgg", MaxResults=1
|
|
)
|
|
assert len(vgg_processing_job_1["ProcessingJobSummaries"]) == 0
|
|
assert vgg_processing_job_1.get("NextToken") is not None
|
|
|
|
vgg_processing_job_6 = sagemaker_client.list_processing_jobs(
|
|
NameContains="vgg", MaxResults=6
|
|
)
|
|
|
|
assert len(vgg_processing_job_6["ProcessingJobSummaries"]) == 1
|
|
assert (
|
|
vgg_processing_job_6["ProcessingJobSummaries"][0]["ProcessingJobName"]
|
|
== "vgg-0"
|
|
)
|
|
assert vgg_processing_job_6.get("NextToken") is not None
|
|
|
|
vgg_processing_job_10 = sagemaker_client.list_processing_jobs(
|
|
NameContains="vgg", MaxResults=10
|
|
)
|
|
|
|
assert len(vgg_processing_job_10["ProcessingJobSummaries"]) == 5
|
|
assert (
|
|
vgg_processing_job_10["ProcessingJobSummaries"][-1]["ProcessingJobName"]
|
|
== "vgg-4"
|
|
)
|
|
assert vgg_processing_job_10.get("NextToken") is None
|
|
|
|
|
|
def test_list_processing_jobs_paginated_with_fragmented_targets(sagemaker_client):
|
|
for i in range(5):
|
|
name = f"xgboost-{i}"
|
|
arn = f"arn:aws:sagemaker:us-east-1:000000000000:x-x/foobar-{i}"
|
|
MyProcessingJobModel(processing_job_name=name, role_arn=arn).save(
|
|
sagemaker_client
|
|
)
|
|
|
|
for i in range(5):
|
|
name = f"vgg-{i}"
|
|
arn = f"arn:aws:sagemaker:us-east-1:000000000000:x-x/barfoo-{i}"
|
|
MyProcessingJobModel(processing_job_name=name, role_arn=arn).save(
|
|
sagemaker_client
|
|
)
|
|
|
|
processing_jobs_with_2 = sagemaker_client.list_processing_jobs(
|
|
NameContains="2", MaxResults=8
|
|
)
|
|
assert len(processing_jobs_with_2["ProcessingJobSummaries"]) == 2
|
|
assert processing_jobs_with_2.get("NextToken") is not None
|
|
|
|
processing_jobs_with_2_next = sagemaker_client.list_processing_jobs(
|
|
NameContains="2",
|
|
MaxResults=1,
|
|
NextToken=processing_jobs_with_2.get("NextToken"),
|
|
)
|
|
assert len(processing_jobs_with_2_next["ProcessingJobSummaries"]) == 0
|
|
assert processing_jobs_with_2_next.get("NextToken") is not None
|
|
|
|
processing_jobs_with_2_next_next = sagemaker_client.list_processing_jobs(
|
|
NameContains="2",
|
|
MaxResults=1,
|
|
NextToken=processing_jobs_with_2_next.get("NextToken"),
|
|
)
|
|
assert len(processing_jobs_with_2_next_next["ProcessingJobSummaries"]) == 0
|
|
assert processing_jobs_with_2_next_next.get("NextToken") is None
|
|
|
|
|
|
def test_add_and_delete_tags_in_training_job(sagemaker_client):
|
|
processing_job_name = "MyProcessingJob"
|
|
role_arn = f"arn:aws:iam::{ACCOUNT_ID}:role/FakeRole"
|
|
container = "382416733822.dkr.ecr.us-east-1.amazonaws.com/linear-learner:1"
|
|
bucket = "my-bucket"
|
|
prefix = "my-prefix"
|
|
app_specification = {
|
|
"ImageUri": container,
|
|
"ContainerEntrypoint": ["python3", "app.py"],
|
|
}
|
|
processing_resources = {
|
|
"ClusterConfig": {
|
|
"InstanceCount": 2,
|
|
"InstanceType": "ml.m5.xlarge",
|
|
"VolumeSizeInGB": 20,
|
|
},
|
|
}
|
|
stopping_condition = {"MaxRuntimeInSeconds": 60 * 60}
|
|
|
|
job = MyProcessingJobModel(
|
|
processing_job_name,
|
|
role_arn,
|
|
container=container,
|
|
bucket=bucket,
|
|
prefix=prefix,
|
|
app_specification=app_specification,
|
|
processing_resources=processing_resources,
|
|
stopping_condition=stopping_condition,
|
|
)
|
|
resp = job.save(sagemaker_client)
|
|
resource_arn = resp["ProcessingJobArn"]
|
|
|
|
tags = [
|
|
{"Key": "myKey", "Value": "myValue"},
|
|
]
|
|
response = sagemaker_client.add_tags(ResourceArn=resource_arn, Tags=tags)
|
|
assert response["ResponseMetadata"]["HTTPStatusCode"] == 200
|
|
|
|
response = sagemaker_client.list_tags(ResourceArn=resource_arn)
|
|
assert response["Tags"] == tags
|
|
|
|
tag_keys = [tag["Key"] for tag in tags]
|
|
response = sagemaker_client.delete_tags(ResourceArn=resource_arn, TagKeys=tag_keys)
|
|
assert response["ResponseMetadata"]["HTTPStatusCode"] == 200
|
|
|
|
response = sagemaker_client.list_tags(ResourceArn=resource_arn)
|
|
assert response["Tags"] == []
|