sagemaker-hyperpod 3.0.2__tar.gz → 3.2.0__tar.gz

{sagemaker_hyperpod-3.0.2/src/sagemaker_hyperpod.egg-info → sagemaker_hyperpod-3.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sagemaker-hyperpod
-Version: 3.0.2
+Version: 3.2.0
 Summary: Amazon SageMaker HyperPod SDK and CLI
 Home-page: https://github.com/aws/sagemaker-hyperpod-cli
 Author: Amazon Web Services
@@ -108,24 +108,13 @@ SageMaker HyperPod CLI currently supports start training job with:
 
 1. Make sure that your local python version is 3.8, 3.9, 3.10 or 3.11.
 
-1. Install ```helm```.
-
-    The SageMaker Hyperpod CLI uses Helm to start training jobs. See also the [Helm installation guide](https://helm.sh/docs/intro/install/).
-
-    ```
-    curl -fsSL -o get_helm.sh https://raw.githubusercontent.com/helm/helm/main/scripts/get-helm-3
-    chmod 700 get_helm.sh
-    ./get_helm.sh
-    rm -f ./get_helm.sh  
-    ```
-
-1. Clone and install the sagemaker-hyperpod-cli package.
+2. Install the sagemaker-hyperpod-cli package.
 
     ```
     pip install sagemaker-hyperpod
     ```
 
-1. Verify if the installation succeeded by running the following command.
+3. Verify if the installation succeeded by running the following command.
 
     ```
     hyp --help
@@ -224,8 +213,15 @@ hyp create hyp-pytorch-job \
     --queue-name "training-queue" \
     --priority "high" \
     --max-retry 3 \
+    --accelerators 8 \
+    --vcpu 96.0 \
+    --memory 1152.0 \
+    --accelerators-limit 8 \
+    --vcpu-limit 96.0 \
+    --memory-limit 1152.0 \
+    --preferred-topology "topology.kubernetes.io/zone=us-west-2a" \
     --volume name=model-data,type=hostPath,mount_path=/data,path=/data \
-    --volume name=training-output,type=pvc,mount_path=/data,claim_name=my-pvc,read_only=false
+    --volume name=training-output,type=pvc,mount_path=/data2,claim_name=my-pvc,read_only=false
 ```
 
 Key required parameters explained:
@@ -246,7 +242,6 @@ hyp create hyp-jumpstart-endpoint \
     --model-id jumpstart-model-id\
     --instance-type ml.g5.8xlarge \
     --endpoint-name endpoint-jumpstart \
-    --tls-output-s3-uri s3://sample-bucket
 ```
 
 
@@ -262,7 +257,7 @@ hyp invoke hyp-jumpstart-endpoint \
 
 ```
 hyp list hyp-jumpstart-endpoint
-hyp get hyp-jumpstart-endpoint --name endpoint-jumpstart
+hyp describe hyp-jumpstart-endpoint --name endpoint-jumpstart
 ```
 
 #### Creating a Custom Inference Endpoint 
@@ -273,7 +268,8 @@ hyp create hyp-custom-endpoint \
     --endpoint-name my-custom-endpoint \
     --model-name my-pytorch-model \
     --model-source-type s3 \
-    --model-location my-pytorch-training/model.tar.gz \
+    --model-location my-pytorch-training \
+    --model-volume-mount-name test-volume \
     --s3-bucket-name your-bucket \
     --s3-region us-east-1 \
     --instance-type ml.g5.8xlarge \
@@ -387,20 +383,17 @@ from sagemaker.hyperpod.inference.config.hp_jumpstart_endpoint_config import Mod
 from sagemaker.hyperpod.inference.hp_jumpstart_endpoint import HPJumpStartEndpoint
 
 model=Model(
-    model_id='deepseek-llm-r1-distill-qwen-1-5b',
-    model_version='2.0.4',
+    model_id='deepseek-llm-r1-distill-qwen-1-5b'
 )
 server=Server(
     instance_type='ml.g5.8xlarge',
 )
 endpoint_name=SageMakerEndpoint(name='<my-endpoint-name>')
-tls_config=TlsConfig(tls_certificate_output_s3_uri='s3://<my-tls-bucket>')
 
 js_endpoint=HPJumpStartEndpoint(
     model=model,
     server=server,
-    sage_maker_endpoint=endpoint_name,
-    tls_config=tls_config,
+    sage_maker_endpoint=endpoint_name
 )
 
 js_endpoint.create()

{sagemaker_hyperpod-3.0.2 → sagemaker_hyperpod-3.2.0}/README.md

@@ -54,24 +54,13 @@ SageMaker HyperPod CLI currently supports start training job with:
 
 1. Make sure that your local python version is 3.8, 3.9, 3.10 or 3.11.
 
-1. Install ```helm```.
-
-    The SageMaker Hyperpod CLI uses Helm to start training jobs. See also the [Helm installation guide](https://helm.sh/docs/intro/install/).
-
-    ```
-    curl -fsSL -o get_helm.sh https://raw.githubusercontent.com/helm/helm/main/scripts/get-helm-3
-    chmod 700 get_helm.sh
-    ./get_helm.sh
-    rm -f ./get_helm.sh  
-    ```
-
-1. Clone and install the sagemaker-hyperpod-cli package.
+2. Install the sagemaker-hyperpod-cli package.
 
     ```
     pip install sagemaker-hyperpod
     ```
 
-1. Verify if the installation succeeded by running the following command.
+3. Verify if the installation succeeded by running the following command.
 
     ```
     hyp --help
@@ -170,8 +159,15 @@ hyp create hyp-pytorch-job \
     --queue-name "training-queue" \
     --priority "high" \
     --max-retry 3 \
+    --accelerators 8 \
+    --vcpu 96.0 \
+    --memory 1152.0 \
+    --accelerators-limit 8 \
+    --vcpu-limit 96.0 \
+    --memory-limit 1152.0 \
+    --preferred-topology "topology.kubernetes.io/zone=us-west-2a" \
     --volume name=model-data,type=hostPath,mount_path=/data,path=/data \
-    --volume name=training-output,type=pvc,mount_path=/data,claim_name=my-pvc,read_only=false
+    --volume name=training-output,type=pvc,mount_path=/data2,claim_name=my-pvc,read_only=false
 ```
 
 Key required parameters explained:
@@ -192,7 +188,6 @@ hyp create hyp-jumpstart-endpoint \
     --model-id jumpstart-model-id\
     --instance-type ml.g5.8xlarge \
     --endpoint-name endpoint-jumpstart \
-    --tls-output-s3-uri s3://sample-bucket
 ```
 
 
@@ -208,7 +203,7 @@ hyp invoke hyp-jumpstart-endpoint \
 
 ```
 hyp list hyp-jumpstart-endpoint
-hyp get hyp-jumpstart-endpoint --name endpoint-jumpstart
+hyp describe hyp-jumpstart-endpoint --name endpoint-jumpstart
 ```
 
 #### Creating a Custom Inference Endpoint 
@@ -219,7 +214,8 @@ hyp create hyp-custom-endpoint \
     --endpoint-name my-custom-endpoint \
     --model-name my-pytorch-model \
     --model-source-type s3 \
-    --model-location my-pytorch-training/model.tar.gz \
+    --model-location my-pytorch-training \
+    --model-volume-mount-name test-volume \
     --s3-bucket-name your-bucket \
     --s3-region us-east-1 \
     --instance-type ml.g5.8xlarge \
@@ -333,20 +329,17 @@ from sagemaker.hyperpod.inference.config.hp_jumpstart_endpoint_config import Mod
 from sagemaker.hyperpod.inference.hp_jumpstart_endpoint import HPJumpStartEndpoint
 
 model=Model(
-    model_id='deepseek-llm-r1-distill-qwen-1-5b',
-    model_version='2.0.4',
+    model_id='deepseek-llm-r1-distill-qwen-1-5b'
 )
 server=Server(
     instance_type='ml.g5.8xlarge',
 )
 endpoint_name=SageMakerEndpoint(name='<my-endpoint-name>')
-tls_config=TlsConfig(tls_certificate_output_s3_uri='s3://<my-tls-bucket>')
 
 js_endpoint=HPJumpStartEndpoint(
     model=model,
     server=server,
-    sage_maker_endpoint=endpoint_name,
-    tls_config=tls_config,
+    sage_maker_endpoint=endpoint_name
 )
 
 js_endpoint.create()

{sagemaker_hyperpod-3.0.2 → sagemaker_hyperpod-3.2.0}/pyproject.toml

@@ -5,7 +5,7 @@ build-backend = "setuptools.build_meta"
 [project]
 dynamic = ["dependencies"]
 name = "sagemaker-hyperpod"
-version = "3.0.2"
+version = "3.2.0"
 description = "Amazon SageMaker HyperPod SDK and CLI"
 readme = "README.md"
 requires-python = ">=3.8"
@@ -112,4 +112,4 @@ docstring-code-format = false
 #
 # This only has an effect when the `docstring-code-format` setting is
 # enabled.
-docstring-code-line-length = "dynamic"
+docstring-code-line-length = "dynamic"

{sagemaker_hyperpod-3.0.2 → sagemaker_hyperpod-3.2.0}/setup.py

@@ -47,7 +47,7 @@ for root, dirs, files in os.walk(
 setup(
     data_files=sagemaker_hyperpod_recipes,
     name="sagemaker-hyperpod",
-    version="3.0.2",
+    version="3.2.0",
     description="Amazon SageMaker HyperPod SDK and CLI",
     long_description=open("README.md").read(),
     long_description_content_type="text/markdown",
@@ -89,7 +89,9 @@ setup(
         "pydantic>=2.10.6,<3.0.0",
         "hyperpod-pytorch-job-template>=1.0.0, <2.0.0",
         "hyperpod-custom-inference-template>=1.0.0, <2.0.0",
-        "hyperpod-jumpstart-inference-template>=1.0.0, <2.0.0"
+         "hyperpod-jumpstart-inference-template>=1.0.0, <2.0.0",
+        # To be enabled after launch
+        #"hyperpod-cluster-stack-template>=1.0.0, <2.0.0"
     ],
     entry_points={
         "console_scripts": [

{sagemaker_hyperpod-3.0.2 → sagemaker_hyperpod-3.2.0}/src/sagemaker/hyperpod/cli/clients/kubernetes_client.py

@@ -51,14 +51,10 @@ class KubernetesClient:
     _instance = None
     _kube_client = None
 
-    def __new__(cls, is_get_capacity: bool = False) -> "KubernetesClient":
+    def __new__(cls, config_file: Optional[str] = None) -> "KubernetesClient":
         if cls._instance is None:
             cls._instance = super(KubernetesClient, cls).__new__(cls)
-            config.load_kube_config(
-                config_file=KUBE_CONFIG_PATH
-                if not is_get_capacity
-                else TEMP_KUBE_CONFIG_FILE
-            )  # or config.load_incluster_config() for in-cluster config
+            config.load_kube_config(config_file=config_file or KUBE_CONFIG_PATH)
             cls._instance._kube_client = client.ApiClient()
         return cls._instance
 

{sagemaker_hyperpod-3.0.2 → sagemaker_hyperpod-3.2.0}/src/sagemaker/hyperpod/cli/commands/cluster.py

@@ -14,8 +14,10 @@ import logging
 import subprocess
 import json
 import sys
+import signal
 import botocore.config
 from collections import defaultdict
+from concurrent.futures import ThreadPoolExecutor, as_completed
 from typing import Any, Dict, List, Optional, Tuple
 
 import boto3
@@ -191,30 +193,33 @@ def list_cluster(
 
     cluster_capacities: List[List[str]] = []
 
-    counter = 0
-    for cluster_name in cluster_names:
-        current_cluster_capacities_size = len(cluster_capacities)
-        rate_limited_operation(
-            cluster_name=cluster_name,
-            validator=validator,
-            sm_client=sm_client,
-            region=region,
-            temp_config_file=TEMP_KUBE_CONFIG_FILE,
-            cluster_capacities=cluster_capacities,
-            namespace=namespace,
-        )
-        # cluster_capacities will only be updated when the cluster
-        # is a valid Hyperpod EKS cluster. This check avoid
-        # we skipped many Hyperpod Slurm clusters and didn't return
-        # any Hyperpod EKS clusters.
-        if len(cluster_capacities) > current_cluster_capacities_size:
-            counter += 1
-        # Currently only support list <= 50 clusters
-        if counter >= 50:
-            logger.debug(
-                "The 'get-clusters' command has reached the maximum number of HyperPod clusters that can be listed, which is 50."
-            )
-            break
+    # Process clusters in parallel with limited concurrency
+    if cluster_names:
+        with ThreadPoolExecutor(max_workers=len(cluster_names)) as executor:
+            futures = {}
+            counter = 0
+
+            for cluster_name in cluster_names[:50]:  # Limit to 50 clusters
+                future = executor.submit(
+                    rate_limited_operation,
+                    cluster_name=cluster_name,
+                    validator=validator,
+                    sm_client=sm_client,
+                    region=region,
+                    temp_config_file=f"{TEMP_KUBE_CONFIG_FILE}_{cluster_name}",
+                    namespace=namespace,
+                )
+                futures[future] = cluster_name
+
+            for future in as_completed(futures):
+                cluster_name = futures[future]
+                try:
+                    result = future.result()
+                    if result:  # Only add if cluster processing was successful
+                        cluster_capacities.extend(result)
+                        counter += 1
+                except Exception as e:
+                    logger.error(f"Error processing cluster {cluster_name}: {e}")
 
     headers = [
         "Cluster",
@@ -245,10 +250,42 @@ def rate_limited_operation(
     sm_client: BaseClient,
     region: Optional[str],
     temp_config_file: str,
-    cluster_capacities: List[List[str]],
     namespace: Optional[List[str]],
-) -> None:
+) -> Optional[List[List[str]]]:
     try:
+        cluster_capacities = []  # Initialize at the beginning
+        
+        # Get cluster details to check instance count
+        cluster_response = sm_client.describe_cluster(ClusterName=cluster_name)
+        cluster_status = cluster_response.get('ClusterStatus', 'Unknown')
+        
+        # Check if cluster has zero instances
+        instance_groups = cluster_response.get('InstanceGroups', [])
+        total_instances = sum(
+            group.get('CurrentCount', 0) for group in instance_groups
+        )
+        
+        # If cluster has 0 instances, add it with 0 nodes
+        if total_instances == 0:
+            logger.info(f"Adding cluster {cluster_name} with 0 instances (status: {cluster_status})")
+            zero_instance_row = [
+                cluster_name,
+                "N/A",  # InstanceType
+                0,      # TotalNodes
+                0,      # AcceleratorDevicesAvailable
+                0,      # NodeHealthStatus=Schedulable
+                "N/A",  # DeepHealthCheckStatus=Passed
+            ]
+            
+            # Add namespace columns with 0 values
+            if namespace:
+                for ns in namespace:
+                    zero_instance_row.extend([0, 0])  # Total and Available accelerator devices
+            
+            cluster_capacities.append(zero_instance_row)
+            return cluster_capacities
+        
+        # Proceed with EKS validation for clusters with instances
         eks_cluster_arn = validator.validate_cluster_and_get_eks_arn(
             cluster_name, sm_client
         )
@@ -256,10 +293,10 @@ def rate_limited_operation(
             logger.warning(
                 f"Cannot find EKS cluster behind {cluster_name}, continue..."
             )
-            return
+            return None
         eks_cluster_name = get_name_from_arn(eks_cluster_arn)
         _update_kube_config(eks_cluster_name, region, temp_config_file)
-        k8s_client = KubernetesClient(is_get_capacity=True)
+        k8s_client = KubernetesClient(config_file=temp_config_file)
         nodes = k8s_client.list_node_with_temp_config(
             temp_config_file, SAGEMAKER_HYPERPOD_NAME_LABEL
         )
@@ -268,25 +305,27 @@ def rate_limited_operation(
         ns_nominal_quota = {}
         ns_quota_usage = {}
 
-        for ns in namespace:
-            sm_managed_namespace = k8s_client.get_sagemaker_managed_namespace(ns)
-            if sm_managed_namespace:
-                quota_allocation_id = sm_managed_namespace.metadata.labels[
-                    SAGEMAKER_QUOTA_ALLOCATION_LABEL
-                ]
-                cluster_queue_name = (
-                    HYPERPOD_NAMESPACE_PREFIX
-                    + quota_allocation_id
-                    + SAGEMAKER_MANAGED_CLUSTER_QUEUE_SUFFIX
-                )
-                cluster_queue = k8s_client.get_cluster_queue(cluster_queue_name)
-                nominal_quota = _get_cluster_queue_nominal_quota(cluster_queue)
-                quota_usage = _get_cluster_queue_quota_usage(cluster_queue)
-                ns_nominal_quota[ns] = nominal_quota
-                ns_quota_usage[ns] = quota_usage
-            else:
-                ns_nominal_quota[ns] = {}
-                ns_quota_usage[ns] = {}
+        if namespace:
+            for ns in namespace:
+                sm_managed_namespace = k8s_client.get_sagemaker_managed_namespace(ns)
+                if sm_managed_namespace:
+                    quota_allocation_id = sm_managed_namespace.metadata.labels[
+                        SAGEMAKER_QUOTA_ALLOCATION_LABEL
+                    ]
+                    cluster_queue_name = (
+                        HYPERPOD_NAMESPACE_PREFIX
+                        + quota_allocation_id
+                        + SAGEMAKER_MANAGED_CLUSTER_QUEUE_SUFFIX
+                    )
+
+                    cluster_queue = k8s_client.get_cluster_queue(cluster_queue_name)
+                    nominal_quota = _get_cluster_queue_nominal_quota(cluster_queue)
+                    quota_usage = _get_cluster_queue_quota_usage(cluster_queue)
+                    ns_nominal_quota[ns] = nominal_quota
+                    ns_quota_usage[ns] = quota_usage
+                else:
+                    ns_nominal_quota[ns] = {}
+                    ns_quota_usage[ns] = {}
 
         for instance_type, nodes_summary in nodes_info.items():
             capacities = [
@@ -297,23 +336,26 @@ def rate_limited_operation(
                 nodes_summary["schedulable"],
                 nodes_summary["deep_health_check_passed"],
             ]
-            for ns in namespace:
-                capacities.append(
-                    ns_nominal_quota.get(ns)
-                    .get(instance_type, {})
-                    .get(NVIDIA_GPU_RESOURCE_LIMIT_KEY, "N/A")
-                )
-                capacities.append(
-                    _get_available_quota(
-                        ns_nominal_quota.get(ns),
-                        ns_quota_usage.get(ns),
-                        instance_type,
-                        NVIDIA_GPU_RESOURCE_LIMIT_KEY,
+            if namespace:
+                for ns in namespace:
+                    capacities.append(
+                        ns_nominal_quota.get(ns)
+                        .get(instance_type, {})
+                        .get(NVIDIA_GPU_RESOURCE_LIMIT_KEY, "N/A")
+                    )
+                    capacities.append(
+                        _get_available_quota(
+                            ns_nominal_quota.get(ns),
+                            ns_quota_usage.get(ns),
+                            instance_type,
+                            NVIDIA_GPU_RESOURCE_LIMIT_KEY,
+                        )
                     )
-                )
             cluster_capacities.append(capacities)
+        return cluster_capacities
     except Exception as e:
         logger.error(f"Error processing cluster {cluster_name}: {e}, continue...")
+        return None
 
 
 def _get_cluster_queue_nominal_quota(cluster_queue):
@@ -519,16 +561,26 @@ def set_cluster_context(
     """
     if debug:
         set_logging_level(logger, logging.DEBUG)
-    validator = ClusterValidator()
-    botocore_config = botocore.config.Config(
-        user_agent_extra=get_user_agent_extra_suffix()
-    )
-    session = boto3.Session(region_name=region) if region else boto3.Session()
-    if not validator.validate_aws_credential(session):
-        logger.error("Cannot connect to HyperPod cluster due to aws credentials error")
-        sys.exit(1)
-
+    
+    timeout = 60  # 1 minute
+    
+    def timeout_handler(signum, frame):
+        raise TimeoutError(f"Operation timed out after {timeout} seconds")
+    
+    # Set up timeout
+    signal.signal(signal.SIGALRM, timeout_handler)
+    signal.alarm(timeout)
+    
     try:
+        validator = ClusterValidator()
+        botocore_config = botocore.config.Config(
+            user_agent_extra=get_user_agent_extra_suffix()
+        )
+        session = boto3.Session(region_name=region) if region else boto3.Session()
+        if not validator.validate_aws_credential(session):
+            logger.error("Cannot connect to HyperPod cluster due to aws credentials error")
+            sys.exit(1)
+
         sm_client = get_sagemaker_client(session, botocore_config)
         hp_cluster_details = sm_client.describe_cluster(ClusterName=cluster_name)
         logger.debug("Fetched hyperpod cluster details")
@@ -542,6 +594,14 @@ def set_cluster_context(
         _update_kube_config(eks_name, region, None)
         k8s_client = KubernetesClient()
         k8s_client.set_context(eks_cluster_arn, namespace)
+        
+        # Cancel the alarm if operation completes successfully
+        signal.alarm(0)
+        logger.info(f"Successfully connected to cluster {cluster_name}")
+        
+    except TimeoutError as e:
+        logger.error("Timed out - Please check credentials, setup configurations  and try again")
+        sys.exit(1)
     except botocore.exceptions.NoRegionError:
         logger.error(
             f"Please ensure you configured AWS default region or use '--region' argument to specify the region"
@@ -552,6 +612,9 @@ def set_cluster_context(
             f"Unexpected error happens when try to connect to cluster {cluster_name}. Error: {e}"
         )
         sys.exit(1)
+    finally:
+        # Ensure alarm is cancelled in all cases
+        signal.alarm(0)
 
 
 @click.command()