Image Spec refactoring and updates (#1525)

nargokul · pintaoz-aws · commit fba3285c421c · 2024-12-04T01:26:49.000-08:00
* Image Spec refactoring and updates

* Unit tests and update function for Image Spec

* Fix hugging face test

* Fix Tests
diff --git a/src/sagemaker/modules/image_spec.py b/src/sagemaker/modules/image_spec.py
@@ -13,41 +13,71 @@
 """ImageSpec class module."""
 from __future__ import absolute_import
 
+import re
+from enum import Enum
 from typing import Optional
 
-from sagemaker import image_uris, Session
-from sagemaker.serverless import ServerlessInferenceConfig
-from sagemaker.training_compiler.config import TrainingCompilerConfig
+from sagemaker import utils
+from sagemaker.image_uris import _validate_version_and_set_if_needed, _version_for_config, \
+    _config_for_framework_and_scope, _validate_py_version_and_set_if_needed, _registry_from_region, ECR_URI_TEMPLATE, \
+    _get_latest_versions, _validate_instance_deprecation, _get_image_tag, _validate_arg
+from packaging.version import Version
+
+DEFAULT_TOLERATE_MODEL = False
+
+
+class Framework(Enum):
+    HUGGING_FACE = "huggingface"
+    HUGGING_FACE_NEURON = "huggingface-neuron"
+    HUGGING_FACE_NEURON_X = "huggingface-neuronx"
+    HUGGING_FACE_LLM = "huggingface-llm"
+    HUGGING_FACE_TEI_GPU = "huggingface-tei"
+    HUGGING_FACE_TEI_CPU = "huggingface-tei-cpu"
+    HUGGING_FACE_LLM_NEURONX = "huggingface-llm-neuronx"
+    HUGGING_FACE_TRAINING_COMPILER = "huggingface-training-compiler"
+    XGBOOST = "xgboost"
+    XG_BOOST_NEO = "xg-boost-neo"
+    SKLEARN = "sklearn"
+    PYTORCH = "pytorch"
+    PYTORCH_TRAINING_COMPILER = "pytorch-training-compiler"
+    DATA_WRANGLER = "data-wrangler"
+    STABILITYAI = "stabilityai"
+    SAGEMAKER_TRITONSERVER = "sagemaker-tritonserver"
+
+
+class ImageScope(Enum):
+    TRAINING = "training"
+    INFERENCE = "inference"
+    INFERENCE_GRAVITON = "inference-graviton"
+
+
+class Processor(Enum):
+    INF = "inf"
+    NEURON = "neuron"
+    GPU = "gpu"
+    CPU = "cpu"
+    TRN = "trn"
 
 
 class ImageSpec:
     """ImageSpec class to get image URI for a specific framework version."""
 
-    def __init__(
-        self,
-        framework_name: str,
-        version: str,
-        image_scope: Optional[str] = None,
-        instance_type: Optional[str] = None,
-        py_version: Optional[str] = None,
-        region: Optional[str] = "us-west-2",
-        accelerator_type: Optional[str] = None,
-        container_version: Optional[str] = None,
-        distribution: Optional[dict] = None,
-        base_framework_version: Optional[str] = None,
-        training_compiler_config: Optional[TrainingCompilerConfig] = None,
-        model_id: Optional[str] = None,
-        model_version: Optional[str] = None,
-        hub_arn: Optional[str] = None,
-        tolerate_vulnerable_model: Optional[bool] = False,
-        tolerate_deprecated_model: Optional[bool] = False,
-        sdk_version: Optional[str] = None,
-        inference_tool: Optional[str] = None,
-        serverless_inference_config: Optional[ServerlessInferenceConfig] = None,
-        config_name: Optional[str] = None,
-        sagemaker_session: Optional[Session] = None,
-    ):
-        self.framework_name = framework_name
+    def __init__(self,
+                 framework: Framework,
+                 processor: Optional[Processor] = Processor.CPU,
+                 region: Optional[str] = "us-west-2",
+                 version=None,
+                 py_version=None,
+                 instance_type=None,
+                 accelerator_type=None,
+                 image_scope: ImageScope = ImageScope.TRAINING,
+                 container_version=None,
+                 distribution=None,
+                 base_framework_version=None,
+                 sdk_version=None,
+                 inference_tool=None):
+        self.framework = framework
+        self.processor = processor
         self.version = version
         self.image_scope = image_scope
         self.instance_type = instance_type
@@ -57,45 +87,175 @@ def __init__(
         self.container_version = container_version
         self.distribution = distribution
         self.base_framework_version = base_framework_version
-        self.training_compiler_config = training_compiler_config
-        self.model_id = model_id
-        self.model_version = model_version
-        self.hub_arn = hub_arn
-        self.tolerate_vulnerable_model = tolerate_vulnerable_model
-        self.tolerate_deprecated_model = tolerate_deprecated_model
         self.sdk_version = sdk_version
         self.inference_tool = inference_tool
-        self.serverless_inference_config = serverless_inference_config
-        self.config_name = config_name
-        self.sagemaker_session = sagemaker_session
-
-    def get_image_uri(
-        self, image_scope: Optional[str] = None, instance_type: Optional[str] = None
-    ) -> str:
-        """Get image URI for a specific framework version."""
-
-        self.image_scope = image_scope or self.image_scope
-        self.instance_type = instance_type or self.instance_type
-        return image_uris.retrieve(
-            framework=self.framework_name,
-            image_scope=self.image_scope,
-            instance_type=self.instance_type,
-            py_version=self.py_version,
-            region=self.region,
-            version=self.version,
-            accelerator_type=self.accelerator_type,
-            container_version=self.container_version,
-            distribution=self.distribution,
-            base_framework_version=self.base_framework_version,
-            training_compiler_config=self.training_compiler_config,
-            model_id=self.model_id,
-            model_version=self.model_version,
-            hub_arn=self.hub_arn,
-            tolerate_vulnerable_model=self.tolerate_vulnerable_model,
-            tolerate_deprecated_model=self.tolerate_deprecated_model,
-            sdk_version=self.sdk_version,
-            inference_tool=self.inference_tool,
-            serverless_inference_config=self.serverless_inference_config,
-            config_name=self.config_name,
-            sagemaker_session=self.sagemaker_session,
-        )
+
+    def update_image_spec(self, **kwargs):
+        for key, value in kwargs.items():
+            if hasattr(self, key):
+                setattr(self, key, value)
+
+    def retrieve(self) -> str:
+        """Retrieves the ECR URI for the Docker image matching the given arguments.
+
+        Ideally this function should not be called directly, rather it should be called from the
+        fit() function inside framework estimator.
+
+        Args:
+            framework (Framework): The name of the framework or algorithm.
+            processor (Processor): The name of the processor (CPU, GPU, etc.).
+            region (str): The AWS region.
+            version (str): The framework or algorithm version. This is required if there is
+                more than one supported version for the given framework or algorithm.
+            py_version (str): The Python version. This is required if there is
+                more than one supported Python version for the given framework version.
+            instance_type (str): The SageMaker instance type. For supported types, see
+                https://aws.amazon.com/sagemaker/pricing. This is required if
+                there are different images for different processor types.
+            accelerator_type (str): Elastic Inference accelerator type. For more, see
+                https://docs.aws.amazon.com/sagemaker/latest/dg/ei.html.
+            image_scope (str): The image type, i.e. what it is used for.
+                Valid values: "training", "inference", "inference_graviton", "eia".
+                If ``accelerator_type`` is set, ``image_scope`` is ignored.
+            container_version (str): the version of docker image.
+                Ideally the value of parameter should be created inside the framework.
+                For custom use, see the list of supported container versions:
+                https://github.com/aws/deep-learning-containers/blob/master/available_images.md
+                (default: None).
+            distribution (dict): A dictionary with information on how to run distributed training
+            sdk_version (str): the version of python-sdk that will be used in the image retrieval.
+                (default: None).
+            inference_tool (str): the tool that will be used to aid in the inference.
+                Valid values: "neuron, neuronx, None"
+                (default: None).
+
+        Returns:
+            str: The ECR URI for the corresponding SageMaker Docker image.
+
+        Raises:
+            NotImplementedError: If the scope is not supported.
+            ValueError: If the combination of arguments specified is not supported or
+                any PipelineVariable object is passed in.
+            VulnerableJumpStartModelError: If any of the dependencies required by the script have
+                known security vulnerabilities.
+            DeprecatedJumpStartModelError: If the version of the model is deprecated.
+        """
+        config = _config_for_framework_and_scope(self.framework.value,
+                                                 self.image_scope.value,
+                                                 self.accelerator_type)
+
+        original_version = self.version
+        try:
+            version = _validate_version_and_set_if_needed(self.version, config, self.framework.value)
+        except ValueError:
+            version = None
+        if not version:
+            version = self._fetch_latest_version_from_config(config)
+
+        version_config = config["versions"][_version_for_config(version, config)]
+
+        if "huggingface" in self.framework.value:
+            if version_config.get("version_aliases"):
+                full_base_framework_version = version_config["version_aliases"].get(
+                    self.base_framework_version, self.base_framework_version
+                )
+            _validate_arg(full_base_framework_version, list(version_config.keys()), "base framework")
+            version_config = version_config.get(full_base_framework_version)
+
+        self.py_version = _validate_py_version_and_set_if_needed(self.py_version,
+                                                            version_config,
+                                                            self.framework.value)
+        version_config = version_config.get(self.py_version) or version_config
+
+        registry = _registry_from_region(self.region, version_config["registries"])
+        endpoint_data = utils._botocore_resolver().construct_endpoint("ecr", self.region)
+        if self.region == "il-central-1" and not endpoint_data:
+            endpoint_data = {"hostname": "ecr.{}.amazonaws.com".format(self.region)}
+        hostname = endpoint_data["hostname"]
+
+        repo = version_config["repository"]
+
+        # if container version is available in .json file, utilize that
+        if version_config.get("container_version"):
+            self.container_version = version_config["container_version"][self.processor.value]
+
+        # Append sdk version in case of trainium instances
+        if repo in ["pytorch-training-neuron"]:
+            if not self.sdk_version:
+                sdk_version = _get_latest_versions(version_config["sdk_versions"])
+            self.container_version = self.sdk_version + "-" + self.container_version
+
+        if self.framework == Framework.HUGGING_FACE:
+            pt_or_tf_version = (
+                re.compile("^(pytorch|tensorflow)(.*)$").match(self.base_framework_version).group(2)
+            )
+            _version = original_version
+
+            if repo in [
+                "huggingface-pytorch-trcomp-training",
+                "huggingface-tensorflow-trcomp-training",
+            ]:
+                _version = version
+            if repo in [
+                "huggingface-pytorch-inference-neuron",
+                "huggingface-pytorch-inference-neuronx",
+            ]:
+                if not sdk_version:
+                    self.sdk_version = _get_latest_versions(version_config["sdk_versions"])
+                self.container_version = self.sdk_version + "-" + self.container_version
+                if config.get("version_aliases").get(original_version):
+                    _version = config.get("version_aliases")[original_version]
+                if (
+                        config.get("versions", {})
+                                .get(_version, {})
+                                .get("version_aliases", {})
+                                .get(self.base_framework_version, {})
+                ):
+                    _base_framework_version = config.get("versions")[_version]["version_aliases"][
+                        self.base_framework_version
+                    ]
+                    pt_or_tf_version = (
+                        re.compile("^(pytorch|tensorflow)(.*)$").match(_base_framework_version).group(2)
+                    )
+
+            tag_prefix = f"{pt_or_tf_version}-transformers{_version}"
+        else:
+            tag_prefix = version_config.get("tag_prefix", version)
+
+        if repo == f"{self.framework.value}-inference-graviton":
+            self.container_version = f"{self.container_version}-sagemaker"
+        _validate_instance_deprecation(self.framework,
+                                       self.instance_type,
+                                       version)
+
+        tag = _get_image_tag(
+            self.container_version,
+            self.distribution,
+            self.image_scope.value,
+            self.framework,
+            self.inference_tool,
+            self.instance_type,
+            self.processor.value,
+            self.py_version,
+            tag_prefix,
+            version)
+
+        if tag:
+            repo += ":{}".format(tag)
+
+        return ECR_URI_TEMPLATE.format(registry=registry, hostname=hostname, repository=repo)
+
+    def _fetch_latest_version_from_config(self,
+                                          framework_config: dict) -> str:
+        if self.image_scope.value in framework_config:
+            if image_scope_config := framework_config[self.image_scope.value]:
+                if version_aliases := image_scope_config["version_aliases"]:
+                    if latest_version := version_aliases["latest"]:
+                        return latest_version
+        versions = list(framework_config["versions"].keys())
+        top_version = versions[0]
+        bottom_version = versions[-1]
+
+        if Version(top_version) >= Version(bottom_version):
+            return top_version
+        return bottom_version
diff --git a/tests/unit/sagemaker/modules/__init__.py b/tests/unit/sagemaker/modules/__init__.py
diff --git a/tests/unit/sagemaker/modules/test_image_spec.py b/tests/unit/sagemaker/modules/test_image_spec.py
@@ -0,0 +1,42 @@
+import unittest
+
+from sagemaker.modules.image_spec import ImageSpec, Framework, Processor
+
+
+class TestImageSpec(unittest.TestCase):
+
+    def test_image_spec_update(self):
+        image_spec = ImageSpec(framework=Framework.HUGGING_FACE)
+        assert image_spec.version == None
+        image_spec.update_image_spec(version="v3")
+        assert image_spec.version == "v3"
+
+    def test_image_spec_retrive(self):
+        # Asserting substrings because full string uri can change with newer versions
+        image_spec = ImageSpec(framework=Framework.XGBOOST)
+        xgboost_uri = image_spec.retrieve()
+        assert "dkr.ecr.us-west-2.amazonaws.com" in xgboost_uri
+        assert "sagemaker-xgboost" in xgboost_uri
+
+        image_spec = ImageSpec(framework=Framework.HUGGING_FACE,
+                               processor=Processor.GPU,
+                               base_framework_version="pytorch2.1.0")
+        hugging_face_uri = image_spec.retrieve()
+        assert "dkr.ecr.us-west-2.amazonaws.com" in hugging_face_uri
+        assert "huggingface-pytorch-training" in hugging_face_uri
+
+        image_spec = ImageSpec(framework=Framework.PYTORCH)
+        pytorch_uri = image_spec.retrieve()
+        assert "dkr.ecr.us-west-2.amazonaws.com" in pytorch_uri
+        assert "pytorch-training" in pytorch_uri
+
+        image_spec = ImageSpec(framework=Framework.SKLEARN)
+        sklearn_uri = image_spec.retrieve()
+        assert "dkr.ecr.us-west-2.amazonaws.com" in sklearn_uri
+        assert "sagemaker-scikit-learn" in sklearn_uri
+
+    def test_image_spec_retrive_with_version(self):
+        image_spec = ImageSpec(framework=Framework.XGBOOST,
+                               version="0.90-1")
+        xgboost_uri = image_spec.retrieve()
+        assert xgboost_uri == '246618743249.dkr.ecr.us-west-2.amazonaws.com/sagemaker-xgboost:0.90-1-cpu-py3'
