Skip to content

change: use recommended inference image uri from Neo API #2923

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Feb 16, 2022
Merged

change: use recommended inference image uri from Neo API #2923

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6d991f2
change: use recommended inference image uri from Neo API
HappyAmazonian Feb 9, 2022
7777e01
Merge branch 'dev' into inf-url
navinsoni Feb 16, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
36 changes: 1 addition & 35 deletions src/sagemaker/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,6 @@
import sagemaker
from sagemaker import (
fw_utils,
image_uris,
local,
s3,
session,
Expand Down Expand Up @@ -657,34 +656,6 @@ def _compilation_job_config(
"job_name": job_name,
}

def _compilation_image_uri(self, region, target_instance_type, framework, framework_version):
"""Retrieve the Neo or Inferentia image URI.

Args:
region (str): The AWS region.
target_instance_type (str): Identifies the device on which you want to run
your model after compilation, for example: ml_c5. For valid values, see
https://docs.aws.amazon.com/sagemaker/latest/dg/API_OutputConfig.html.
framework (str): The framework name.
framework_version (str): The framework version.
"""
framework_prefix = ""
framework_suffix = ""

if framework == "xgboost":
framework_suffix = "-neo"
elif target_instance_type.startswith("ml_inf"):
framework_prefix = "inferentia-"
else:
framework_prefix = "neo-"

return image_uris.retrieve(
"{}{}{}".format(framework_prefix, framework, framework_suffix),
region,
instance_type=target_instance_type,
version=framework_version,
)

def package_for_edge(
self,
output_path,
Expand Down Expand Up @@ -849,12 +820,7 @@ def compile(
if target_instance_family == "ml_eia2":
pass
elif target_instance_family.startswith("ml_"):
self.image_uri = self._compilation_image_uri(
self.sagemaker_session.boto_region_name,
target_instance_family,
framework,
framework_version,
)
self.image_uri = job_status.get("InferenceImage", None)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this param expected to be there always? What if this is None? Is the removed method needed?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Neo service puts InferenceImage in the response according to the API response. In the case that the user uses ml instance, the InferenceImage will not be None from our service.
We are making this change so the user can get the latest inference url recommendation.

Is the removed method needed?
Thanks for pointing out. Will delete it

self._is_compiled_model = True
else:
LOGGER.warning(
Expand Down
73 changes: 38 additions & 35 deletions tests/unit/sagemaker/model/test_neo.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,12 +20,15 @@
MODEL_DATA = "s3://bucket/model.tar.gz"
MODEL_IMAGE = "mi"

IMAGE_URI = "inference-container-uri"

REGION = "us-west-2"

NEO_REGION_ACCOUNT = "301217895009"
DESCRIBE_COMPILATION_JOB_RESPONSE = {
"CompilationJobStatus": "Completed",
"ModelArtifacts": {"S3ModelArtifacts": "s3://output-path/model.tar.gz"},
"InferenceImage": IMAGE_URI,
}


Expand All @@ -52,12 +55,7 @@ def test_compile_model_for_inferentia(sagemaker_session):
framework_version="1.15.0",
job_name="compile-model",
)
assert (
"{}.dkr.ecr.{}.amazonaws.com/sagemaker-neo-tensorflow:1.15.0-inf-py3".format(
NEO_REGION_ACCOUNT, REGION
)
== model.image_uri
)
assert DESCRIBE_COMPILATION_JOB_RESPONSE["InferenceImage"] == model.image_uri
assert model._is_compiled_model is True


Expand Down Expand Up @@ -271,11 +269,12 @@ def test_deploy_add_compiled_model_suffix_to_endpoint_name_from_model_name(sagem
assert model.endpoint_name.startswith("{}-ml-c4".format(model_name))


@patch("sagemaker.session.Session")
def test_compile_with_framework_version_15(session):
session.return_value.boto_region_name = REGION
def test_compile_with_framework_version_15(sagemaker_session):
sagemaker_session.wait_for_compilation_job = Mock(
return_value=DESCRIBE_COMPILATION_JOB_RESPONSE
)

model = _create_model()
model = _create_model(sagemaker_session)
model.compile(
target_instance_family="ml_c4",
input_shape={"data": [1, 3, 1024, 1024]},
Expand All @@ -286,14 +285,15 @@ def test_compile_with_framework_version_15(session):
job_name="compile-model",
)

assert "1.5" in model.image_uri
assert IMAGE_URI == model.image_uri


@patch("sagemaker.session.Session")
def test_compile_with_framework_version_16(session):
session.return_value.boto_region_name = REGION
def test_compile_with_framework_version_16(sagemaker_session):
sagemaker_session.wait_for_compilation_job = Mock(
return_value=DESCRIBE_COMPILATION_JOB_RESPONSE
)

model = _create_model()
model = _create_model(sagemaker_session)
model.compile(
target_instance_family="ml_c4",
input_shape={"data": [1, 3, 1024, 1024]},
Expand All @@ -304,26 +304,7 @@ def test_compile_with_framework_version_16(session):
job_name="compile-model",
)

assert "1.6" in model.image_uri


@patch("sagemaker.session.Session")
def test_compile_validates_framework_version(session):
session.return_value.boto_region_name = REGION

model = _create_model()
with pytest.raises(ValueError) as e:
model.compile(
target_instance_family="ml_c4",
input_shape={"data": [1, 3, 1024, 1024]},
output_path="s3://output",
role="role",
framework="pytorch",
framework_version="1.6.1",
job_name="compile-model",
)

assert "Unsupported neo-pytorch version: 1.6.1." in str(e)
assert IMAGE_URI == model.image_uri


@patch("sagemaker.session.Session")
Expand All @@ -347,3 +328,25 @@ def test_compile_with_pytorch_neo_in_ml_inf(session):
)
!= model.image_uri
)


def test_compile_validates_framework_version(sagemaker_session):
sagemaker_session.wait_for_compilation_job = Mock(
return_value={
"CompilationJobStatus": "Completed",
"ModelArtifacts": {"S3ModelArtifacts": "s3://output-path/model.tar.gz"},
"InferenceImage": None,
}
)
model = _create_model(sagemaker_session)
model.compile(
target_instance_family="ml_c4",
input_shape={"data": [1, 3, 1024, 1024]},
output_path="s3://output",
role="role",
framework="pytorch",
framework_version="1.6.1",
job_name="compile-model",
)

assert model.image_uri is None
15 changes: 7 additions & 8 deletions tests/unit/test_mxnet.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,6 +68,8 @@

ENV_INPUT = {"env_key1": "env_val1", "env_key2": "env_val2", "env_key3": "env_val3"}

INFERENCE_IMAGE_URI = "inference-uri"


@pytest.fixture()
def sagemaker_session():
Expand All @@ -83,7 +85,10 @@ def sagemaker_session():
)

describe = {"ModelArtifacts": {"S3ModelArtifacts": "s3://m/m.tar.gz"}}
describe_compilation = {"ModelArtifacts": {"S3ModelArtifacts": "s3://m/model_c5.tar.gz"}}
describe_compilation = {
"ModelArtifacts": {"S3ModelArtifacts": "s3://m/model_c5.tar.gz"},
"InferenceImage": INFERENCE_IMAGE_URI,
}
session.sagemaker_client.create_model_package.side_effect = MODEL_PKG_RESPONSE
session.sagemaker_client.describe_training_job = Mock(return_value=describe)
session.sagemaker_client.describe_endpoint = Mock(return_value=ENDPOINT_DESC)
Expand Down Expand Up @@ -195,12 +200,6 @@ def _create_compilation_job(input_shape, output_location):
}


def _neo_inference_image(mxnet_version):
return "301217895009.dkr.ecr.us-west-2.amazonaws.com/sagemaker-inference-{}:{}-cpu-py3".format(
FRAMEWORK.lower(), mxnet_version
)


@patch("sagemaker.estimator.name_from_base")
@patch("sagemaker.utils.create_tar_file", MagicMock())
def test_create_model(
Expand Down Expand Up @@ -422,7 +421,7 @@ def test_mxnet_neo(time, strftime, sagemaker_session, neo_mxnet_version):
actual_compile_model_args = sagemaker_session.method_calls[3][2]
assert expected_compile_model_args == actual_compile_model_args

assert compiled_model.image_uri == _neo_inference_image(neo_mxnet_version)
assert compiled_model.image_uri == INFERENCE_IMAGE_URI

predictor = mx.deploy(1, CPU, use_compiled_model=True)
assert isinstance(predictor, MXNetPredictor)
Expand Down