aws-sdk-sagemaker 1.406.0 → 1.407.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: d0e34856cbdcb87283d1d242622a7407ce18763cc513b2f03fcba4675ae0e578
-  data.tar.gz: 907774d87667d808bb9cffa813a7382a45321ff0a10c3c5a3c05c1799a83c5b4
+  metadata.gz: 577c772d7fb6bbccc01a1001ac5a449b15342c22334a6c38afd6e61f74b563b5
+  data.tar.gz: a184b5ed3f54d960c7e8d89b09267c8f15ee8d2ad3e0f6e43b2f557a8b31d7ad
 SHA512:
-  metadata.gz: 4a37c322f52be2723c2e7bb87338e9594890921269c40d50082ad32b2bfce86ddf466b17555ebf99832483d031c9677703076e6991dedd07f6354b6f95f9e0eb
-  data.tar.gz: c763f57d75fc78a76d5b7f0278d3b2dab9a70b66e710602cae5692aa8127aec1eb37fd39cac76aff21b5679803d67e064b33432d349cca3e1a0749ebe72b4c38
+  metadata.gz: 2495fa23bf6edd84fa24b25808d6baba0a82732b9ed538443772d4f40e7a0ce6a61820a23f2e54663fbb15d900e6d72b152688b38b6b9a9039239ae3e6703e26
+  data.tar.gz: f7ea44076206308bb05080c70343392ff0714c69660dc418ea8d977b60172d9cdd8d50178ee57cba7764bf5ca29211b1b0ec22035d93aae3aaced676faae29d9

data/CHANGELOG.md

@@ -1,6 +1,11 @@
 Unreleased Changes
 ------------------
 
+1.407.0 (2026-10-30)
+------------------
+
+* Feature - Create endpoints with AcceleratorPartitionLayouts on ProductionVariant and deploy inference components with AcceleratorPartitionType in ComputeResourceRequirements to divide accelerators into hardware-isolated NVIDIA MIG partitions, running multiple models per GPU. P4d, P5, G7, G7e.
+
 1.406.0 (2026-10-29)
 ------------------
 

data/VERSION

@@ -1 +1 @@
-1.406.0
+1.407.0

data/lib/aws-sdk-sagemaker/client.rb

@@ -4911,6 +4911,11 @@ module Aws::SageMaker
     #           capacity_reservation_preference: "capacity-reservations-only", # accepts capacity-reservations-only
     #           ml_reservation_arn: "MlReservationArn",
     #         },
+    #         accelerator_partition_layouts: [
+    #           {
+    #             accelerator_partition_types: ["AcceleratorPartitionType"],
+    #           },
+    #         ],
     #       },
     #     ],
     #     data_capture_config: {
@@ -5033,6 +5038,11 @@ module Aws::SageMaker
     #           capacity_reservation_preference: "capacity-reservations-only", # accepts capacity-reservations-only
     #           ml_reservation_arn: "MlReservationArn",
     #         },
+    #         accelerator_partition_layouts: [
+    #           {
+    #             accelerator_partition_types: ["AcceleratorPartitionType"],
+    #           },
+    #         ],
     #       },
     #     ],
     #     execution_role_arn: "RoleArn",
@@ -6500,6 +6510,7 @@ module Aws::SageMaker
     #         number_of_accelerator_devices_required: 1.0,
     #         min_memory_required_in_mb: 1, # required
     #         max_memory_required_in_mb: 1,
+    #         accelerator_partition_type: "AcceleratorPartitionType",
     #       },
     #       base_inference_component_name: "InferenceComponentName",
     #       data_cache_config: {
@@ -6541,6 +6552,7 @@ module Aws::SageMaker
     #           number_of_accelerator_devices_required: 1.0,
     #           min_memory_required_in_mb: 1, # required
     #           max_memory_required_in_mb: 1,
+    #           accelerator_partition_type: "AcceleratorPartitionType",
     #         },
     #         base_inference_component_name: "InferenceComponentName",
     #         data_cache_config: {
@@ -16652,6 +16664,9 @@ module Aws::SageMaker
     #   resp.production_variants[0].capacity_reservation_config.ec2_capacity_reservations[0].total_instance_count #=> Integer
     #   resp.production_variants[0].capacity_reservation_config.ec2_capacity_reservations[0].available_instance_count #=> Integer
     #   resp.production_variants[0].capacity_reservation_config.ec2_capacity_reservations[0].used_by_current_endpoint #=> Integer
+    #   resp.production_variants[0].accelerator_partition_layouts #=> Array
+    #   resp.production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types #=> Array
+    #   resp.production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types[0] #=> String
     #   resp.data_capture_config.enable_capture #=> Boolean
     #   resp.data_capture_config.capture_status #=> String, one of "Started", "Stopped"
     #   resp.data_capture_config.current_sampling_percentage #=> Integer
@@ -16720,6 +16735,9 @@ module Aws::SageMaker
     #   resp.pending_deployment_summary.production_variants[0].routing_config.routing_strategy #=> String, one of "LEAST_OUTSTANDING_REQUESTS", "RANDOM", "PREFIX_AWARE"
     #   resp.pending_deployment_summary.production_variants[0].routing_config.prefix_aware_routing_config.prefix_length #=> Integer
     #   resp.pending_deployment_summary.production_variants[0].routing_config.prefix_aware_routing_config.concurrency_threshold #=> Integer
+    #   resp.pending_deployment_summary.production_variants[0].accelerator_partition_layouts #=> Array
+    #   resp.pending_deployment_summary.production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types #=> Array
+    #   resp.pending_deployment_summary.production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types[0] #=> String
     #   resp.pending_deployment_summary.start_time #=> Time
     #   resp.pending_deployment_summary.shadow_production_variants #=> Array
     #   resp.pending_deployment_summary.shadow_production_variants[0].variant_name #=> String
@@ -16755,6 +16773,9 @@ module Aws::SageMaker
     #   resp.pending_deployment_summary.shadow_production_variants[0].routing_config.routing_strategy #=> String, one of "LEAST_OUTSTANDING_REQUESTS", "RANDOM", "PREFIX_AWARE"
     #   resp.pending_deployment_summary.shadow_production_variants[0].routing_config.prefix_aware_routing_config.prefix_length #=> Integer
     #   resp.pending_deployment_summary.shadow_production_variants[0].routing_config.prefix_aware_routing_config.concurrency_threshold #=> Integer
+    #   resp.pending_deployment_summary.shadow_production_variants[0].accelerator_partition_layouts #=> Array
+    #   resp.pending_deployment_summary.shadow_production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types #=> Array
+    #   resp.pending_deployment_summary.shadow_production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types[0] #=> String
     #   resp.explainer_config.clarify_explainer_config.enable_explanations #=> String
     #   resp.explainer_config.clarify_explainer_config.inference_config.features_attribute #=> String
     #   resp.explainer_config.clarify_explainer_config.inference_config.content_template #=> String
@@ -16820,6 +16841,9 @@ module Aws::SageMaker
     #   resp.shadow_production_variants[0].capacity_reservation_config.ec2_capacity_reservations[0].total_instance_count #=> Integer
     #   resp.shadow_production_variants[0].capacity_reservation_config.ec2_capacity_reservations[0].available_instance_count #=> Integer
     #   resp.shadow_production_variants[0].capacity_reservation_config.ec2_capacity_reservations[0].used_by_current_endpoint #=> Integer
+    #   resp.shadow_production_variants[0].accelerator_partition_layouts #=> Array
+    #   resp.shadow_production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types #=> Array
+    #   resp.shadow_production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types[0] #=> String
     #   resp.metrics_config.enable_enhanced_metrics #=> Boolean
     #   resp.metrics_config.enable_detailed_observability #=> Boolean
     #   resp.metrics_config.metric_publish_frequency_in_seconds #=> Integer
@@ -16904,6 +16928,9 @@ module Aws::SageMaker
     #   resp.production_variants[0].inference_ami_version #=> String, one of "al2-ami-sagemaker-inference-gpu-2", "al2-ami-sagemaker-inference-gpu-2-1", "al2-ami-sagemaker-inference-gpu-3-1", "al2-ami-sagemaker-inference-neuron-2", "al2023-ami-sagemaker-inference-gpu-4-1"
     #   resp.production_variants[0].capacity_reservation_config.capacity_reservation_preference #=> String, one of "capacity-reservations-only"
     #   resp.production_variants[0].capacity_reservation_config.ml_reservation_arn #=> String
+    #   resp.production_variants[0].accelerator_partition_layouts #=> Array
+    #   resp.production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types #=> Array
+    #   resp.production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types[0] #=> String
     #   resp.data_capture_config.enable_capture #=> Boolean
     #   resp.data_capture_config.initial_sampling_percentage #=> Integer
     #   resp.data_capture_config.destination_s3_uri #=> String
@@ -16980,6 +17007,9 @@ module Aws::SageMaker
     #   resp.shadow_production_variants[0].inference_ami_version #=> String, one of "al2-ami-sagemaker-inference-gpu-2", "al2-ami-sagemaker-inference-gpu-2-1", "al2-ami-sagemaker-inference-gpu-3-1", "al2-ami-sagemaker-inference-neuron-2", "al2023-ami-sagemaker-inference-gpu-4-1"
     #   resp.shadow_production_variants[0].capacity_reservation_config.capacity_reservation_preference #=> String, one of "capacity-reservations-only"
     #   resp.shadow_production_variants[0].capacity_reservation_config.ml_reservation_arn #=> String
+    #   resp.shadow_production_variants[0].accelerator_partition_layouts #=> Array
+    #   resp.shadow_production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types #=> Array
+    #   resp.shadow_production_variants[0].accelerator_partition_layouts[0].accelerator_partition_types[0] #=> String
     #   resp.execution_role_arn #=> String
     #   resp.vpc_config.security_group_ids #=> Array
     #   resp.vpc_config.security_group_ids[0] #=> String
@@ -17918,6 +17948,7 @@ module Aws::SageMaker
     #   resp.specification.compute_resource_requirements.number_of_accelerator_devices_required #=> Float
     #   resp.specification.compute_resource_requirements.min_memory_required_in_mb #=> Integer
     #   resp.specification.compute_resource_requirements.max_memory_required_in_mb #=> Integer
+    #   resp.specification.compute_resource_requirements.accelerator_partition_type #=> String
     #   resp.specification.base_inference_component_name #=> String
     #   resp.specification.data_cache_config.enable_caching #=> Boolean
     #   resp.specification.scheduling_config.placement_strategy #=> String, one of "SPREAD", "BINPACK"
@@ -17941,6 +17972,7 @@ module Aws::SageMaker
     #   resp.specifications[0].compute_resource_requirements.number_of_accelerator_devices_required #=> Float
     #   resp.specifications[0].compute_resource_requirements.min_memory_required_in_mb #=> Integer
     #   resp.specifications[0].compute_resource_requirements.max_memory_required_in_mb #=> Integer
+    #   resp.specifications[0].compute_resource_requirements.accelerator_partition_type #=> String
     #   resp.specifications[0].base_inference_component_name #=> String
     #   resp.specifications[0].data_cache_config.enable_caching #=> Boolean
     #   resp.specifications[0].scheduling_config.placement_strategy #=> String, one of "SPREAD", "BINPACK"
@@ -32812,6 +32844,7 @@ module Aws::SageMaker
     #         number_of_accelerator_devices_required: 1.0,
     #         min_memory_required_in_mb: 1, # required
     #         max_memory_required_in_mb: 1,
+    #         accelerator_partition_type: "AcceleratorPartitionType",
     #       },
     #       base_inference_component_name: "InferenceComponentName",
     #       data_cache_config: {
@@ -32853,6 +32886,7 @@ module Aws::SageMaker
     #           number_of_accelerator_devices_required: 1.0,
     #           min_memory_required_in_mb: 1, # required
     #           max_memory_required_in_mb: 1,
+    #           accelerator_partition_type: "AcceleratorPartitionType",
     #         },
     #         base_inference_component_name: "InferenceComponentName",
     #         data_cache_config: {
@@ -35120,7 +35154,7 @@ module Aws::SageMaker
         tracer: tracer
       )
       context[:gem_name] = 'aws-sdk-sagemaker'
-      context[:gem_version] = '1.406.0'
+      context[:gem_version] = '1.407.0'
       Seahorse::Client::Request.new(handlers, context)
     end
 

data/lib/aws-sdk-sagemaker/client_api.rb

@@ -92,6 +92,10 @@ module Aws::SageMaker
     AbsoluteBorrowLimitResourceList = Shapes::ListShape.new(name: 'AbsoluteBorrowLimitResourceList')
     AcceleratorPartitionConfig = Shapes::StructureShape.new(name: 'AcceleratorPartitionConfig')
     AcceleratorPartitionConfigCountInteger = Shapes::IntegerShape.new(name: 'AcceleratorPartitionConfigCountInteger')
+    AcceleratorPartitionLayout = Shapes::StructureShape.new(name: 'AcceleratorPartitionLayout')
+    AcceleratorPartitionLayouts = Shapes::ListShape.new(name: 'AcceleratorPartitionLayouts')
+    AcceleratorPartitionType = Shapes::StringShape.new(name: 'AcceleratorPartitionType')
+    AcceleratorPartitionTypeList = Shapes::ListShape.new(name: 'AcceleratorPartitionTypeList')
     AcceleratorsAmount = Shapes::IntegerShape.new(name: 'AcceleratorsAmount')
     Accept = Shapes::StringShape.new(name: 'Accept')
     AcceptEula = Shapes::BooleanShape.new(name: 'AcceptEula')
@@ -2345,6 +2349,7 @@ module Aws::SageMaker
     ProductId = Shapes::StringShape.new(name: 'ProductId')
     ProductListings = Shapes::ListShape.new(name: 'ProductListings')
     ProductionVariant = Shapes::StructureShape.new(name: 'ProductionVariant')
+    ProductionVariantAcceleratorPartitionLayoutsList = Shapes::ListShape.new(name: 'ProductionVariantAcceleratorPartitionLayoutsList')
     ProductionVariantAcceleratorType = Shapes::StringShape.new(name: 'ProductionVariantAcceleratorType')
     ProductionVariantCapacityReservationConfig = Shapes::StructureShape.new(name: 'ProductionVariantCapacityReservationConfig')
     ProductionVariantCapacityReservationSummary = Shapes::StructureShape.new(name: 'ProductionVariantCapacityReservationSummary')
@@ -3324,6 +3329,13 @@ module Aws::SageMaker
     AcceleratorPartitionConfig.add_member(:count, Shapes::ShapeRef.new(shape: AcceleratorPartitionConfigCountInteger, required: true, location_name: "Count", metadata: {"box" => true}))
     AcceleratorPartitionConfig.struct_class = Types::AcceleratorPartitionConfig
 
+    AcceleratorPartitionLayout.add_member(:accelerator_partition_types, Shapes::ShapeRef.new(shape: AcceleratorPartitionTypeList, location_name: "AcceleratorPartitionTypes"))
+    AcceleratorPartitionLayout.struct_class = Types::AcceleratorPartitionLayout
+
+    AcceleratorPartitionLayouts.member = Shapes::ShapeRef.new(shape: AcceleratorPartitionLayout)
+
+    AcceleratorPartitionTypeList.member = Shapes::ShapeRef.new(shape: AcceleratorPartitionType)
+
     ActionSource.add_member(:source_uri, Shapes::ShapeRef.new(shape: SourceUri, required: true, location_name: "SourceUri"))
     ActionSource.add_member(:source_type, Shapes::ShapeRef.new(shape: String256, location_name: "SourceType"))
     ActionSource.add_member(:source_id, Shapes::ShapeRef.new(shape: String256, location_name: "SourceId"))
@@ -8557,6 +8569,7 @@ module Aws::SageMaker
     InferenceComponentComputeResourceRequirements.add_member(:number_of_accelerator_devices_required, Shapes::ShapeRef.new(shape: NumberOfAcceleratorDevices, location_name: "NumberOfAcceleratorDevicesRequired"))
     InferenceComponentComputeResourceRequirements.add_member(:min_memory_required_in_mb, Shapes::ShapeRef.new(shape: MemoryInMb, required: true, location_name: "MinMemoryRequiredInMb"))
     InferenceComponentComputeResourceRequirements.add_member(:max_memory_required_in_mb, Shapes::ShapeRef.new(shape: MemoryInMb, location_name: "MaxMemoryRequiredInMb"))
+    InferenceComponentComputeResourceRequirements.add_member(:accelerator_partition_type, Shapes::ShapeRef.new(shape: AcceleratorPartitionType, location_name: "AcceleratorPartitionType"))
     InferenceComponentComputeResourceRequirements.struct_class = Types::InferenceComponentComputeResourceRequirements
 
     InferenceComponentContainerSpecification.add_member(:image, Shapes::ShapeRef.new(shape: ContainerImage, location_name: "Image"))
@@ -11230,6 +11243,7 @@ module Aws::SageMaker
     PendingProductionVariantSummary.add_member(:desired_serverless_config, Shapes::ShapeRef.new(shape: ProductionVariantServerlessConfig, location_name: "DesiredServerlessConfig"))
     PendingProductionVariantSummary.add_member(:managed_instance_scaling, Shapes::ShapeRef.new(shape: ProductionVariantManagedInstanceScaling, location_name: "ManagedInstanceScaling"))
     PendingProductionVariantSummary.add_member(:routing_config, Shapes::ShapeRef.new(shape: ProductionVariantRoutingConfig, location_name: "RoutingConfig"))
+    PendingProductionVariantSummary.add_member(:accelerator_partition_layouts, Shapes::ShapeRef.new(shape: AcceleratorPartitionLayouts, location_name: "AcceleratorPartitionLayouts"))
     PendingProductionVariantSummary.struct_class = Types::PendingProductionVariantSummary
 
     PendingProductionVariantSummaryList.member = Shapes::ShapeRef.new(shape: PendingProductionVariantSummary)
@@ -11515,8 +11529,11 @@ module Aws::SageMaker
     ProductionVariant.add_member(:routing_config, Shapes::ShapeRef.new(shape: ProductionVariantRoutingConfig, location_name: "RoutingConfig"))
     ProductionVariant.add_member(:inference_ami_version, Shapes::ShapeRef.new(shape: ProductionVariantInferenceAmiVersion, location_name: "InferenceAmiVersion"))
     ProductionVariant.add_member(:capacity_reservation_config, Shapes::ShapeRef.new(shape: ProductionVariantCapacityReservationConfig, location_name: "CapacityReservationConfig"))
+    ProductionVariant.add_member(:accelerator_partition_layouts, Shapes::ShapeRef.new(shape: ProductionVariantAcceleratorPartitionLayoutsList, location_name: "AcceleratorPartitionLayouts"))
     ProductionVariant.struct_class = Types::ProductionVariant
 
+    ProductionVariantAcceleratorPartitionLayoutsList.member = Shapes::ShapeRef.new(shape: AcceleratorPartitionLayout)
+
     ProductionVariantCapacityReservationConfig.add_member(:capacity_reservation_preference, Shapes::ShapeRef.new(shape: CapacityReservationPreference, location_name: "CapacityReservationPreference"))
     ProductionVariantCapacityReservationConfig.add_member(:ml_reservation_arn, Shapes::ShapeRef.new(shape: MlReservationArn, location_name: "MlReservationArn"))
     ProductionVariantCapacityReservationConfig.struct_class = Types::ProductionVariantCapacityReservationConfig
@@ -11579,6 +11596,7 @@ module Aws::SageMaker
     ProductionVariantSummary.add_member(:managed_instance_scaling, Shapes::ShapeRef.new(shape: ProductionVariantManagedInstanceScaling, location_name: "ManagedInstanceScaling"))
     ProductionVariantSummary.add_member(:routing_config, Shapes::ShapeRef.new(shape: ProductionVariantRoutingConfig, location_name: "RoutingConfig"))
     ProductionVariantSummary.add_member(:capacity_reservation_config, Shapes::ShapeRef.new(shape: ProductionVariantCapacityReservationSummary, location_name: "CapacityReservationConfig"))
+    ProductionVariantSummary.add_member(:accelerator_partition_layouts, Shapes::ShapeRef.new(shape: AcceleratorPartitionLayouts, location_name: "AcceleratorPartitionLayouts"))
     ProductionVariantSummary.struct_class = Types::ProductionVariantSummary
 
     ProductionVariantSummaryList.member = Shapes::ShapeRef.new(shape: ProductionVariantSummary)

data/lib/aws-sdk-sagemaker/types.rb

@@ -883,6 +883,32 @@ module Aws::SageMaker
       include Aws::Structure
     end
 
+    # A layout of accelerator partitions for a production variant. On NVIDIA
+    # GPUs, the partitions are Multi-Instance GPU (MIG) instances. SageMaker
+    # applies the layout to every accelerator on the variant's instance
+    # type. You can't change the layout of a running endpoint. To use a
+    # different layout, create a new endpoint.
+    #
+    # @!attribute [rw] accelerator_partition_types
+    #   The list of partition profiles to carve on each accelerator that
+    #   this layout applies to. On NVIDIA GPUs these are Multi-Instance GPU
+    #   (MIG) profiles, for example `["mig-3g.40gb", "mig-4g.40gb"]`. The
+    #   order of the profiles doesn't change how the accelerator is
+    #   partitioned. However, when you update an endpoint, a layout with the
+    #   same profiles in a different order counts as a different layout, and
+    #   the update is rejected. Repeat a profile to request more than one
+    #   partition of that size: for example, `["mig-3g.40gb",
+    #   "mig-3g.40gb"]` requests two `mig-3g.40gb` partitions.
+    #   @return [Array<String>]
+    #
+    # @see http://docs.aws.amazon.com/goto/WebAPI/sagemaker-2017-07-24/AcceleratorPartitionLayout AWS API Documentation
+    #
+    class AcceleratorPartitionLayout < Struct.new(
+      :accelerator_partition_types)
+      SENSITIVE = []
+      include Aws::Structure
+    end
+
     # A structure describing the source of an action.
     #
     # @!attribute [rw] source_uri
@@ -31884,13 +31910,25 @@ module Aws::SageMaker
     #   to an inference component.
     #   @return [Integer]
     #
+    # @!attribute [rw] accelerator_partition_type
+    #   The accelerator partition profile this inference component occupies;
+    #   on NVIDIA GPUs this is a Multi-Instance GPU (MIG) profile such as
+    #   `mig-3g.40gb`. It must be one of the partition profiles declared in
+    #   the target endpoint's accelerator partition layout, and when set,
+    #   `NumberOfAcceleratorDevicesRequired` must be `1`. This field is
+    #   required when the endpoint's production variant has an accelerator
+    #   partition layout. For endpoints without a layout, omit this field to
+    #   place the inference component on a whole accelerator.
+    #   @return [String]
+    #
     # @see http://docs.aws.amazon.com/goto/WebAPI/sagemaker-2017-07-24/InferenceComponentComputeResourceRequirements AWS API Documentation
     #
     class InferenceComponentComputeResourceRequirements < Struct.new(
       :number_of_cpu_cores_required,
       :number_of_accelerator_devices_required,
       :min_memory_required_in_mb,
-      :max_memory_required_in_mb)
+      :max_memory_required_in_mb,
+      :accelerator_partition_type)
       SENSITIVE = []
       include Aws::Structure
     end
@@ -46302,6 +46340,14 @@ module Aws::SageMaker
     #   the instances that the endpoint hosts.
     #   @return [Types::ProductionVariantRoutingConfig]
     #
+    # @!attribute [rw] accelerator_partition_layouts
+    #   The accelerator partition layout that SageMaker is applying to this
+    #   production variant. On NVIDIA GPUs, the profiles are Multi-Instance
+    #   GPU (MIG) profiles. SageMaker returns the accelerator partition
+    #   layout as you provided it in the pending deployment configuration,
+    #   without normalization.
+    #   @return [Array<Types::AcceleratorPartitionLayout>]
+    #
     # @see http://docs.aws.amazon.com/goto/WebAPI/sagemaker-2017-07-24/PendingProductionVariantSummary AWS API Documentation
     #
     class PendingProductionVariantSummary < Struct.new(
@@ -46318,7 +46364,8 @@ module Aws::SageMaker
       :current_serverless_config,
       :desired_serverless_config,
       :managed_instance_scaling,
-      :routing_config)
+      :routing_config,
+      :accelerator_partition_layouts)
       SENSITIVE = []
       include Aws::Structure
     end
@@ -47097,17 +47144,37 @@ module Aws::SageMaker
     #     characters from the text content of the messages array.
     #
     #   The endpoint routes requests that share the same prefix to the same
-    #   instance. Set this value to cover shared content (such as system
-    #   prompts) plus enough unique content to distribute workloads across
-    #   instances.
+    #   instance and, when that instance hosts more than one copy of the
+    #   invoked inference component, to the same copy. Set this value to
+    #   cover shared content (such as system prompts) plus enough unique
+    #   content to distribute workloads across instances and copies.
     #   @return [Integer]
     #
     # @!attribute [rw] concurrency_threshold
-    #   The maximum number of in-flight requests on the target instance
-    #   before the endpoint routes to another instance. Required when
-    #   `RoutingStrategy` is `PREFIX_AWARE`. When in-flight requests on the
-    #   prefix-selected instance reach this threshold, the endpoint routes
-    #   the request to an instance with more available capacity.
+    #   The number of in-flight requests on the selected instance or
+    #   inference component copy at which the endpoint routes a request to
+    #   another copy or instance. Required when `RoutingStrategy` is
+    #   `PREFIX_AWARE`.
+    #
+    #   On an endpoint without inference components, the endpoint compares
+    #   this threshold with the number of in-flight requests on the selected
+    #   instance. When that number reaches the threshold, the endpoint
+    #   routes the request to another instance.
+    #
+    #   On an endpoint with inference components, the endpoint compares this
+    #   threshold with the number of in-flight requests on the selected copy
+    #   of the invoked inference component. Requests to other inference
+    #   components on the same instance don't count. When the instance
+    #   hosts more than one copy of the inference component, the endpoint
+    #   routes requests that share a prefix to the same copy and, while that
+    #   copy has reached the threshold, to another copy on the instance that
+    #   is under the threshold. When every copy on the instance has reached
+    #   the threshold, the endpoint routes the request to another instance.
+    #
+    #   The threshold applies to every instance and every copy of every
+    #   inference component in the production variant, so choose a value
+    #   that suits the instance or copy that can serve the fewest concurrent
+    #   requests.
     #   @return [Integer]
     #
     # @see http://docs.aws.amazon.com/goto/WebAPI/sagemaker-2017-07-24/PrefixAwareRoutingConfig AWS API Documentation
@@ -47934,6 +48001,17 @@ module Aws::SageMaker
     #   SageMaker AI reserves for an endpoint.
     #   @return [Types::ProductionVariantCapacityReservationConfig]
     #
+    # @!attribute [rw] accelerator_partition_layouts
+    #   The accelerator partition layout for the accelerators backing this
+    #   production variant. A layout lists the partition profiles to carve
+    #   on the accelerators it applies to; on NVIDIA GPUs these are
+    #   Multi-Instance GPU (MIG) profiles such as `mig-3g.40gb`. Specify a
+    #   single layout: SageMaker applies it uniformly to every accelerator
+    #   on the variant's instance type, and you place one inference
+    #   component on each partition. Omit this field to use whole,
+    #   unpartitioned accelerators.
+    #   @return [Array<Types::AcceleratorPartitionLayout>]
+    #
     # @see http://docs.aws.amazon.com/goto/WebAPI/sagemaker-2017-07-24/ProductionVariant AWS API Documentation
     #
     class ProductionVariant < Struct.new(
@@ -47954,7 +48032,8 @@ module Aws::SageMaker
       :managed_instance_scaling,
       :routing_config,
       :inference_ami_version,
-      :capacity_reservation_config)
+      :capacity_reservation_config,
+      :accelerator_partition_layouts)
       SENSITIVE = []
       include Aws::Structure
     end
@@ -48189,10 +48268,13 @@ module Aws::SageMaker
     #     instance.
     #
     #   * `PREFIX_AWARE`: The endpoint routes requests that share the same
-    #     prompt prefix to the same instance. When the number of in-flight
-    #     requests on the selected instance reaches the configured
-    #     threshold, the endpoint routes the request to an instance with
-    #     more available capacity.
+    #     prompt prefix to the same instance and, when that instance hosts
+    #     more than one copy of the invoked inference component, to the same
+    #     copy. When the number of in-flight requests on the selected
+    #     instance or copy reaches the `ConcurrencyThreshold` value, the
+    #     endpoint routes the request to another copy on that instance or to
+    #     another instance. The endpoint applies prefix-aware routing only
+    #     when two or more instances can serve the request.
     #   @return [String]
     #
     # @!attribute [rw] prefix_aware_routing_config
@@ -48377,6 +48459,14 @@ module Aws::SageMaker
     #   SageMaker AI reserves for an endpoint.
     #   @return [Types::ProductionVariantCapacityReservationSummary]
     #
+    # @!attribute [rw] accelerator_partition_layouts
+    #   The accelerator partition layout that SageMaker applies to this
+    #   production variant. On NVIDIA GPUs, the profiles are Multi-Instance
+    #   GPU (MIG) profiles. SageMaker returns the accelerator partition
+    #   layout as you provided it when you created the variant, without
+    #   normalization.
+    #   @return [Array<Types::AcceleratorPartitionLayout>]
+    #
     # @see http://docs.aws.amazon.com/goto/WebAPI/sagemaker-2017-07-24/ProductionVariantSummary AWS API Documentation
     #
     class ProductionVariantSummary < Struct.new(
@@ -48392,7 +48482,8 @@ module Aws::SageMaker
       :desired_serverless_config,
       :managed_instance_scaling,
       :routing_config,
-      :capacity_reservation_config)
+      :capacity_reservation_config,
+      :accelerator_partition_layouts)
       SENSITIVE = []
       include Aws::Structure
     end
@@ -52155,24 +52246,8 @@ module Aws::SageMaker
     # @!attribute [rw] sequence_length
     #   The maximum sequence length, in tokens, that the customization job
     #   supports. SageMaker uses this value to select a training
-    #   configuration for the base model that you specify. The parameter
-    #   supports the following values:
-    #
-    #   * `1K`
-    #
-    #   * `2K`
-    #
-    #   * `4K`
-    #
-    #   * `8K`
-    #
-    #   * `16K`
-    #
-    #   * `32K`
-    #
-    #   * `64K`
-    #
-    #   * `128K`
+    #   configuration for the base model that you specify. The value must be
+    #   an integer value from `1K` to `1024K` (for example, `32K`).
     #
     #   If you don't specify a value, SageMaker selects a training
     #   configuration based on the other values that you specify. The

data/lib/aws-sdk-sagemaker.rb

@@ -55,7 +55,7 @@ module Aws::SageMaker
   autoload :EndpointProvider, 'aws-sdk-sagemaker/endpoint_provider'
   autoload :Endpoints, 'aws-sdk-sagemaker/endpoints'
 
-  GEM_VERSION = '1.406.0'
+  GEM_VERSION = '1.407.0'
 
 end
 

data/sig/params.rbs

@@ -828,7 +828,12 @@ module Aws
       capacity_reservation_config: {
         capacity_reservation_preference: ("capacity-reservations-only")?,
         ml_reservation_arn: ::String?
-      }?
+      }?,
+      accelerator_partition_layouts: Array[
+        {
+          accelerator_partition_types: Array[::String]?
+        }
+      ]?
     }
 
       type feature_definition = {
@@ -968,7 +973,8 @@ module Aws
         number_of_cpu_cores_required: ::Float?,
         number_of_accelerator_devices_required: ::Float?,
         min_memory_required_in_mb: ::Integer,
-        max_memory_required_in_mb: ::Integer?
+        max_memory_required_in_mb: ::Integer?,
+        accelerator_partition_type: ::String?
       }?,
       base_inference_component_name: ::String?,
       data_cache_config: {

data/sig/types.rbs

@@ -271,6 +271,11 @@ module Aws::SageMaker
       SENSITIVE: []
     end
 
+    class AcceleratorPartitionLayout
+      attr_accessor accelerator_partition_types: ::Array[::String]
+      SENSITIVE: []
+    end
+
     class ActionSource
       attr_accessor source_uri: ::String
       attr_accessor source_type: ::String
@@ -6683,6 +6688,7 @@ module Aws::SageMaker
       attr_accessor number_of_accelerator_devices_required: ::Float
       attr_accessor min_memory_required_in_mb: ::Integer
       attr_accessor max_memory_required_in_mb: ::Integer
+      attr_accessor accelerator_partition_type: ::String
       SENSITIVE: []
     end
 
@@ -9978,6 +9984,7 @@ module Aws::SageMaker
       attr_accessor desired_serverless_config: Types::ProductionVariantServerlessConfig
       attr_accessor managed_instance_scaling: Types::ProductionVariantManagedInstanceScaling
       attr_accessor routing_config: Types::ProductionVariantRoutingConfig
+      attr_accessor accelerator_partition_layouts: ::Array[Types::AcceleratorPartitionLayout]
       SENSITIVE: []
     end
 
@@ -10294,6 +10301,7 @@ module Aws::SageMaker
       attr_accessor routing_config: Types::ProductionVariantRoutingConfig
       attr_accessor inference_ami_version: ("al2-ami-sagemaker-inference-gpu-2" | "al2-ami-sagemaker-inference-gpu-2-1" | "al2-ami-sagemaker-inference-gpu-3-1" | "al2-ami-sagemaker-inference-neuron-2" | "al2023-ami-sagemaker-inference-gpu-4-1")
       attr_accessor capacity_reservation_config: Types::ProductionVariantCapacityReservationConfig
+      attr_accessor accelerator_partition_layouts: ::Array[Types::AcceleratorPartitionLayout]
       SENSITIVE: []
     end
 
@@ -10374,6 +10382,7 @@ module Aws::SageMaker
       attr_accessor managed_instance_scaling: Types::ProductionVariantManagedInstanceScaling
       attr_accessor routing_config: Types::ProductionVariantRoutingConfig
       attr_accessor capacity_reservation_config: Types::ProductionVariantCapacityReservationSummary
+      attr_accessor accelerator_partition_layouts: ::Array[Types::AcceleratorPartitionLayout]
       SENSITIVE: []
     end
 

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: aws-sdk-sagemaker
 version: !ruby/object:Gem::Version
-  version: 1.406.0
+  version: 1.407.0
 platform: ruby
 authors:
 - Amazon Web Services