feat: local download dir for Model and Estimator classes (#3602)

Author: evakravi

src/sagemaker/fw_utils.py

@@ -422,7 +422,20 @@ def tar_and_upload_dir(
     script_name = script if directory else os.path.basename(script)
     dependencies = dependencies or []
     key = "%s/sourcedir.tar.gz" % s3_key_prefix
-    tmp = tempfile.mkdtemp()
+    if (
+        settings is not None
+        and settings.local_download_dir is not None
+        and not (
+            os.path.exists(settings.local_download_dir)
+            and os.path.isdir(settings.local_download_dir)
+        )
+    ):
+        raise ValueError(
+            "Inputted directory for storing newly generated temporary directory does "
+            f"not exist: '{settings.local_download_dir}'"
+        )
+    local_download_dir = None if settings is None else settings.local_download_dir
+    tmp = tempfile.mkdtemp(dir=local_download_dir)
     encrypt_artifact = True if settings is None else settings.encrypt_repacked_artifacts
 
     try:

src/sagemaker/session_settings.py

@@ -18,17 +18,25 @@
 class SessionSettings(object):
     """Optional container class for settings to apply to a SageMaker session."""
 
-    def __init__(self, encrypt_repacked_artifacts=True) -> None:
+    def __init__(self, encrypt_repacked_artifacts=True, local_download_dir=None) -> None:
         """Initialize the ``SessionSettings`` of a SageMaker ``Session``.
 
         Args:
             encrypt_repacked_artifacts (bool): Flag to indicate whether to encrypt the artifacts
                 at rest in S3 using the default AWS managed KMS key for S3 when a custom KMS key
                 is not provided (Default: True).
+            local_download_dir (str): Optional. A path specifying the local directory
+                for downloading artifacts. (Default: None).
         """
         self._encrypt_repacked_artifacts = encrypt_repacked_artifacts
+        self._local_download_dir = local_download_dir
 
     @property
     def encrypt_repacked_artifacts(self) -> bool:
         """Return True if repacked artifacts at rest in S3 should be encrypted by default."""
         return self._encrypt_repacked_artifacts
+
+    @property
+    def local_download_dir(self) -> str:
+        """Return path specifying the local directory for downloading artifacts."""
+        return self._local_download_dir

src/sagemaker/utils.py

@@ -358,7 +358,7 @@ def create_tar_file(source_files, target=None):
 
 
 @contextlib.contextmanager
-def _tmpdir(suffix="", prefix="tmp"):
+def _tmpdir(suffix="", prefix="tmp", directory=None):
     """Create a temporary directory with a context manager.
 
     The file is deleted when the context exits.
@@ -369,11 +369,18 @@ def _tmpdir(suffix="", prefix="tmp"):
             suffix, otherwise there will be no suffix.
         prefix (str): If prefix is specified, the file name will begin with that
             prefix; otherwise, a default prefix is used.
+        directory (str): If a directory is specified, the file will be downloaded
+            in this directory; otherwise, a default directory is used.
 
     Returns:
         str: path to the directory
     """
-    tmp = tempfile.mkdtemp(suffix=suffix, prefix=prefix, dir=None)
+    if directory is not None and not (os.path.exists(directory) and os.path.isdir(directory)):
+        raise ValueError(
+            "Inputted directory for storing newly generated temporary "
+            f"directory does not exist: '{directory}'"
+        )
+    tmp = tempfile.mkdtemp(suffix=suffix, prefix=prefix, dir=directory)
     yield tmp
     shutil.rmtree(tmp)
 
@@ -427,7 +434,13 @@ def repack_model(
     """
     dependencies = dependencies or []
 
-    with _tmpdir() as tmp:
+    local_download_dir = (
+        None
+        if sagemaker_session.settings is None
+        or sagemaker_session.settings.local_download_dir is None
+        else sagemaker_session.settings.local_download_dir
+    )
+    with _tmpdir(directory=local_download_dir) as tmp:
         model_dir = _extract_model(model_uri, sagemaker_session, tmp)
 
         _create_or_update_code_dir(

tests/unit/sagemaker/automl/test_auto_ml.py

@@ -18,6 +18,7 @@
 from mock import Mock, patch
 from sagemaker import AutoML, AutoMLJob, AutoMLInput, CandidateEstimator, PipelineModel
 from sagemaker.predictor import Predictor
+from sagemaker.session_settings import SessionSettings
 from sagemaker.workflow.functions import Join
 
 MODEL_DATA = "s3://bucket/model.tar.gz"
@@ -254,6 +255,7 @@ def sagemaker_session():
         boto_region_name=REGION,
         config=None,
         local_mode=False,
+        settings=SessionSettings(),
     )
     sms.default_bucket = Mock(name="default_bucket", return_value=BUCKET_NAME)
     sms.upload_data = Mock(name="upload_data", return_value=DEFAULT_S3_INPUT_DATA)

tests/unit/sagemaker/huggingface/test_estimator.py

@@ -20,6 +20,8 @@
 from mock import MagicMock, Mock, patch
 
 from sagemaker.huggingface import HuggingFace, HuggingFaceModel
+from sagemaker.session_settings import SessionSettings
+
 
 from .huggingface_utils import get_full_gpu_image_uri, GPU_INSTANCE_TYPE, REGION
 
@@ -63,6 +65,7 @@ def fixture_sagemaker_session():
         local_mode=False,
         s3_resource=None,
         s3_client=None,
+        settings=SessionSettings(),
     )
 
     describe = {"ModelArtifacts": {"S3ModelArtifacts": "s3://m/m.tar.gz"}}

tests/unit/sagemaker/huggingface/test_processing.py

@@ -17,6 +17,7 @@
 
 from sagemaker.huggingface.processing import HuggingFaceProcessor
 from sagemaker.fw_utils import UploadedCode
+from sagemaker.session_settings import SessionSettings
 
 from .huggingface_utils import get_full_gpu_image_uri, GPU_INSTANCE_TYPE, REGION
 
@@ -42,6 +43,7 @@ def sagemaker_session():
         boto_region_name=REGION,
         config=None,
         local_mode=False,
+        settings=SessionSettings(),
     )
     session_mock.default_bucket = Mock(name="default_bucket", return_value=BUCKET_NAME)
 

tests/unit/sagemaker/image_uris/jumpstart/conftest.py

@@ -15,6 +15,8 @@
 from mock.mock import Mock
 import pytest
 
+from sagemaker.session_settings import SessionSettings
+
 REGION_NAME = "us-west-2"
 BUCKET_NAME = "some-bucket-name"
 
@@ -26,6 +28,7 @@ def session():
         boto_session=boto_mock,
         boto_region_name=REGION_NAME,
         config=None,
+        settings=SessionSettings(),
     )
     sms.default_bucket = Mock(return_value=BUCKET_NAME)
     return sms

tests/unit/sagemaker/local/test_local_utils.py

@@ -17,6 +17,7 @@
 from mock import patch, Mock
 
 import sagemaker.local.utils
+from sagemaker.session_settings import SessionSettings
 
 
 @patch("sagemaker.local.utils.os.path")
@@ -42,7 +43,9 @@ def test_move_to_destination_local(recursive_copy):
 @patch("shutil.rmtree", Mock())
 @patch("sagemaker.local.utils.recursive_copy")
 def test_move_to_destination_s3(recursive_copy):
-    sms = Mock()
+    sms = Mock(
+        settings=SessionSettings(),
+    )
 
     # without trailing slash in prefix
     sagemaker.local.utils.move_to_destination("/tmp/data", "s3://bucket/path", "job", sms)

tests/unit/sagemaker/model/test_framework_model.py

@@ -21,6 +21,8 @@
 import pytest
 from mock import MagicMock, Mock, patch
 
+from sagemaker.session_settings import SessionSettings
+
 MODEL_DATA = "s3://bucket/model.tar.gz"
 MODEL_IMAGE = "mi"
 ENTRY_POINT = "blah.py"
@@ -89,6 +91,7 @@ def sagemaker_session():
         local_mode=False,
         s3_client=None,
         s3_resource=None,
+        settings=SessionSettings(),
     )
     sms.default_bucket = Mock(name="default_bucket", return_value=BUCKET_NAME)
     return sms

tests/unit/sagemaker/model/test_model.py

@@ -776,3 +776,27 @@ def test_register_calls_model_package_args(get_model_package_args, sagemaker_ses
         == get_model_package_args.call_args_list[0][1]["validation_specification"]
     ), """ValidationSpecification from model.register method is not identical to validation_spec from
          get_model_package_args"""
+
+
+@patch("sagemaker.utils.repack_model")
+def test_model_local_download_dir(repack_model, sagemaker_session):
+
+    source_dir = "s3://blah/blah/blah"
+    local_download_dir = "local download dir"
+
+    sagemaker_session.settings.local_download_dir = local_download_dir
+
+    t = Model(
+        entry_point=ENTRY_POINT_INFERENCE,
+        role=ROLE,
+        sagemaker_session=sagemaker_session,
+        source_dir=source_dir,
+        image_uri=IMAGE_URI,
+        model_data=MODEL_DATA,
+    )
+    t.deploy(instance_type=INSTANCE_TYPE, initial_instance_count=INSTANCE_COUNT)
+
+    assert (
+        repack_model.call_args_list[0][1]["sagemaker_session"].settings.local_download_dir
+        == local_download_dir
+    )

tests/unit/sagemaker/spark/test_processing.py

@@ -18,6 +18,7 @@
 import pytest
 
 from sagemaker.processing import ProcessingInput, ProcessingOutput
+from sagemaker.session_settings import SessionSettings
 from sagemaker.spark.processing import (
     PySparkProcessor,
     SparkJarProcessor,
@@ -57,6 +58,7 @@ def sagemaker_session():
         boto_region_name=REGION,
         config=None,
         local_mode=False,
+        settings=SessionSettings(),
     )
     session_mock.default_bucket = Mock(name="default_bucket", return_value=BUCKET_NAME)
 

tests/unit/sagemaker/tensorflow/test_estimator.py

@@ -21,6 +21,7 @@
 import pytest
 
 from sagemaker.estimator import _TrainingJob
+from sagemaker.session_settings import SessionSettings
 from sagemaker.tensorflow import TensorFlow
 from sagemaker.instance_group import InstanceGroup
 from sagemaker.workflow.parameters import ParameterString, ParameterBoolean
@@ -71,6 +72,7 @@ def sagemaker_session():
         local_mode=False,
         s3_resource=None,
         s3_client=None,
+        settings=SessionSettings(),
     )
     session.default_bucket = Mock(name="default_bucket", return_value=BUCKET_NAME)
     session.expand_role = Mock(name="expand_role", return_value=ROLE)

tests/unit/sagemaker/training_compiler/test_huggingface_pytorch_compiler.py

@@ -24,6 +24,7 @@
 from sagemaker.huggingface import HuggingFace, TrainingCompilerConfig
 from sagemaker.huggingface.model import HuggingFaceModel
 from sagemaker.instance_group import InstanceGroup
+from sagemaker.session_settings import SessionSettings
 
 from tests.unit.sagemaker.training_compiler import EC2_GPU_INSTANCE_CLASSES
 
@@ -72,6 +73,7 @@ def fixture_sagemaker_session():
         local_mode=False,
         s3_resource=None,
         s3_client=None,
+        settings=SessionSettings(),
     )
 
     describe = {"ModelArtifacts": {"S3ModelArtifacts": "s3://m/m.tar.gz"}}

tests/unit/sagemaker/training_compiler/test_huggingface_tensorflow_compiler.py

@@ -22,6 +22,7 @@
 from sagemaker import image_uris
 from sagemaker.huggingface import HuggingFace, TrainingCompilerConfig
 from sagemaker.huggingface.model import HuggingFaceModel
+from sagemaker.session_settings import SessionSettings
 
 from tests.unit.sagemaker.training_compiler import EC2_GPU_INSTANCE_CLASSES
 
@@ -70,6 +71,7 @@ def fixture_sagemaker_session():
         local_mode=False,
         s3_resource=None,
         s3_client=None,
+        settings=SessionSettings(),
     )
 
     describe = {"ModelArtifacts": {"S3ModelArtifacts": "s3://m/m.tar.gz"}}

tests/unit/sagemaker/training_compiler/test_pytorch_compiler.py

@@ -24,6 +24,7 @@
 from sagemaker.pytorch import PyTorch, TrainingCompilerConfig
 from sagemaker.pytorch.model import PyTorchModel
 from sagemaker.instance_group import InstanceGroup
+from sagemaker.session_settings import SessionSettings
 
 from tests.unit.sagemaker.training_compiler import EC2_GPU_INSTANCE_CLASSES
 
@@ -71,6 +72,7 @@ def fixture_sagemaker_session():
         local_mode=False,
         s3_resource=None,
         s3_client=None,
+        settings=SessionSettings(),
     )
 
     describe = {"ModelArtifacts": {"S3ModelArtifacts": "s3://m/m.tar.gz"}}

tests/unit/sagemaker/training_compiler/test_tensorflow_compiler.py

@@ -21,6 +21,7 @@
 from mock import MagicMock, Mock, patch
 
 from sagemaker import image_uris
+from sagemaker.session_settings import SessionSettings
 from sagemaker.tensorflow import TensorFlow, TrainingCompilerConfig
 
 from tests.unit.sagemaker.training_compiler import EC2_GPU_INSTANCE_CLASSES
@@ -76,6 +77,7 @@ def fixture_sagemaker_session():
         local_mode=False,
         s3_resource=None,
         s3_client=None,
+        settings=SessionSettings(),
     )
 
     describe = {"ModelArtifacts": {"S3ModelArtifacts": "s3://m/m.tar.gz"}}

tests/unit/sagemaker/wrangler/test_processing.py

@@ -14,6 +14,7 @@
 
 import pytest
 from mock import Mock, MagicMock
+from sagemaker.session_settings import SessionSettings
 
 from sagemaker.wrangler.processing import DataWranglerProcessor
 from sagemaker.processing import ProcessingInput
@@ -36,6 +37,7 @@ def sagemaker_session():
         boto_region_name=REGION,
         config=None,
         local_mode=False,
+        settings=SessionSettings(),
     )
     session_mock.expand_role.return_value = ROLE
     return session_mock

tests/unit/test_amazon_estimator.py

@@ -23,6 +23,7 @@
     _build_shards,
     FileSystemRecordSet,
 )
+from sagemaker.session_settings import SessionSettings
 
 COMMON_ARGS = {"role": "myrole", "instance_count": 1, "instance_type": "ml.c4.xlarge"}
 
@@ -40,6 +41,7 @@ def sagemaker_session():
         region_name=REGION,
         config=None,
         local_mode=False,
+        settings=SessionSettings(),
     )
     sms.boto_region_name = REGION
     sms.default_bucket = Mock(name="default_bucket", return_value=BUCKET_NAME)

tests/unit/test_analytics.py

@@ -24,6 +24,7 @@
     HyperparameterTuningJobAnalytics,
     TrainingJobAnalytics,
 )
+from sagemaker.session_settings import SessionSettings
 
 BUCKET_NAME = "mybucket"
 REGION = "us-west-2"
@@ -47,6 +48,7 @@ def create_sagemaker_session(
         boto_region_name=REGION,
         config=None,
         local_mode=False,
+        settings=SessionSettings(),
     )
     sms.default_bucket = Mock(name="default_bucket", return_value=BUCKET_NAME)
     sms.sagemaker_client.describe_hyper_parameter_tuning_job = Mock(

tests/unit/test_chainer.py

@@ -24,6 +24,7 @@
 from sagemaker.chainer import defaults
 from sagemaker.chainer import Chainer
 from sagemaker.chainer import ChainerPredictor, ChainerModel
+from sagemaker.session_settings import SessionSettings
 
 DATA_DIR = os.path.join(os.path.dirname(__file__), "..", "data")
 SCRIPT_PATH = os.path.join(DATA_DIR, "dummy_script.py")
@@ -62,6 +63,7 @@ def sagemaker_session():
         local_mode=False,
         s3_resource=None,
         s3_client=None,
+        settings=SessionSettings(),
     )
 
     describe = {"ModelArtifacts": {"S3ModelArtifacts": "s3://m/m.tar.gz"}}