Skip to content

Benchmark feature fixes tests #3

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

Jump to bottom
Merged
merged 31 commits into from
May 2, 2024
Merged

Benchmark feature fixes tests #3

merged 31 commits into from
May 2, 2024

Conversation

makungaj1
Copy link
Owner

@makungaj1 makungaj1 commented May 1, 2024 •
edited
Loading

Issue #, if available:

Description of changes:

  • Unit tests
  • Customize Python LRU to only cache when output is complete

Testing done:
Notebook

-- Without identity-based policy that allows the pricing:GetProducts action

jumpstart_model = JumpStartModel(model_id="meta-textgeneration-llama-3-70b")
jumpstart_model.list_deployment_configs()

Instance rate metrics will be omitted. Reason: User: arn:aws:sts::312206380606:assumed-role/AmazonSageMaker-ExecutionRole-20230707T131628/SageMaker is not authorized to perform: pricing:GetProducts because no identity-based policy allows the pricing:GetProducts action
[{'DeploymentConfigName': 'tgi',
  'DeploymentArgs': {'ImageUri': '875423407011.dkr.ecr.us-west-2.amazonaws.com/suzuka-beta:djl-serving-v3.2',
   'ModelData': {'S3DataSource': {'S3Uri': 's3://jumpstart-private-cache-prod-us-west-2/meta-textgeneration/meta-textgeneration-llama-3-70b/artifacts/inference-prepack/v1.0.0/',
     'S3DataType': 'S3Prefix',
     'CompressionType': 'None'}},
   'Environment': {'SAGEMAKER_PROGRAM': 'inference.py',
    'ENDPOINT_SERVER_TIMEOUT': '3600',
    'MODEL_CACHE_ROOT': '/opt/ml/model',
    'SAGEMAKER_ENV': '1',
    'HF_MODEL_ID': '/opt/ml/model',
    'SERVING_LOAD_MODELS': 'test::MPI=/opt/ml/model',
    'OPTION_MODEL_ID': '/opt/ml/model',
    'OPTION_SPECULATIVE_DRAFT_MODEL': 'sagemaker',
    'OPTION_TENSOR_PARALLEL_DEGREE': 'max',
    'OPTION_MAX_ROLLING_BATCH_SIZE': '64',
    'OPTION_ROLLING_BATCH': 'lmi-dist',
    'OPTION_GPU_MEMORY_UTILIZATION': '0.85',
    'SAGEMAKER_MODEL_SERVER_WORKERS': '1'},
   'InstanceType': 'ml.p4d.24xlarge',
   'ComputeResourceRequirements': {'MinMemoryRequiredInMb': 589824,
    'NumberOfAcceleratorDevicesRequired': 8},
   'ModelDataDownloadTimeout': 1200,
   'ContainerStartupHealthCheckTimeout': 1200},
  'AccelerationConfigs': None,
  'BenchmarkMetrics': None},
 {'DeploymentConfigName': 'lmi',
  'DeploymentArgs': {'ImageUri': '875423407011.dkr.ecr.us-west-2.amazonaws.com/suzuka-beta:djl-serving-v3.2',
   'ModelData': {'S3DataSource': {'S3Uri': 's3://jumpstart-private-cache-prod-us-west-2/meta-textgeneration/meta-textgeneration-llama-3-70b/artifacts/inference-prepack/v1.0.0/',
     'S3DataType': 'S3Prefix',
     'CompressionType': 'None'}},
   'Environment': {'SAGEMAKER_PROGRAM': 'inference.py',
    'ENDPOINT_SERVER_TIMEOUT': '3600',
    'MODEL_CACHE_ROOT': '/opt/ml/model',
    'SAGEMAKER_ENV': '1',
    'HF_MODEL_ID': '/opt/ml/model',
    'SERVING_LOAD_MODELS': 'test::MPI=/opt/ml/model',
    'OPTION_MODEL_ID': '/opt/ml/model',
    'OPTION_SPECULATIVE_DRAFT_MODEL': 'sagemaker',
    'OPTION_TENSOR_PARALLEL_DEGREE': 'max',
    'OPTION_MAX_ROLLING_BATCH_SIZE': '64',
    'OPTION_ROLLING_BATCH': 'lmi-dist',
    'OPTION_GPU_MEMORY_UTILIZATION': '0.85',
    'SAGEMAKER_MODEL_SERVER_WORKERS': '1'},
   'InstanceType': 'ml.p4d.24xlarge',
   'ComputeResourceRequirements': {'MinMemoryRequiredInMb': 589824,
    'NumberOfAcceleratorDevicesRequired': 8},
   'ModelDataDownloadTimeout': 1200,
   'ContainerStartupHealthCheckTimeout': 1200},
  'AccelerationConfigs': None,
  'BenchmarkMetrics': {'ml.p4d.24xlarge': [{'name': 'Min Latency',
     'value': '26',
     'unit': 'ms/token'},
    {'name': 'Max Throughput', 'value': '983', 'unit': 'tokens/sec'},
    {'name': 'Max Concurrency', 'value': '256', 'unit': 'count'}]}},
 {'DeploymentConfigName': 'lmi-accelerated',
  'DeploymentArgs': {'ImageUri': '875423407011.dkr.ecr.us-west-2.amazonaws.com/suzuka-beta:djl-serving-v3.2',
   'ModelData': {'S3DataSource': {'S3Uri': 's3://jumpstart-private-cache-prod-us-west-2/meta-textgeneration/meta-textgeneration-llama-3-70b/artifacts/inference-prepack/v1.0.0/',
     'S3DataType': 'S3Prefix',
     'CompressionType': 'None'}},
   'Environment': {'SAGEMAKER_PROGRAM': 'inference.py',
    'ENDPOINT_SERVER_TIMEOUT': '3600',
    'MODEL_CACHE_ROOT': '/opt/ml/model',
    'SAGEMAKER_ENV': '1',
    'HF_MODEL_ID': '/opt/ml/model',
    'SERVING_LOAD_MODELS': 'test::MPI=/opt/ml/model',
    'OPTION_MODEL_ID': '/opt/ml/model',
    'OPTION_SPECULATIVE_DRAFT_MODEL': 'sagemaker',
    'OPTION_TENSOR_PARALLEL_DEGREE': 'max',
    'OPTION_MAX_ROLLING_BATCH_SIZE': '64',
    'OPTION_ROLLING_BATCH': 'lmi-dist',
    'OPTION_GPU_MEMORY_UTILIZATION': '0.85',
    'SAGEMAKER_MODEL_SERVER_WORKERS': '1'},
   'InstanceType': 'ml.p4d.24xlarge',
   'ComputeResourceRequirements': {'MinMemoryRequiredInMb': 589824,
    'NumberOfAcceleratorDevicesRequired': 8},
   'ModelDataDownloadTimeout': 1200,
   'ContainerStartupHealthCheckTimeout': 1200},
  'AccelerationConfigs': None,
  'BenchmarkMetrics': {'ml.p4d.24xlarge': [{'name': 'Min Latency',
     'value': '10',
     'unit': 'ms/token'},
    {'name': 'Max Throughput', 'value': '635', 'unit': 'tokens/sec'},
    {'name': 'Max Concurrency', 'value': '64', 'unit': 'count'}]}}]

jumpstart_model.display_benchmark_metrics()

Instance rate metrics will be omitted. Reason: User: arn:aws:sts::312206380606:assumed-role/AmazonSageMaker-ExecutionRole-20230707T131628/SageMaker is not authorized to perform: pricing:GetProducts because no identity-based policy allows the pricing:GetProducts action
| Instance Type             | Config Name     |   Min Latency (ms/token) |   Max Throughput (tokens/sec) |   Max Concurrency (count) |
|:--------------------------|:----------------|-------------------------:|------------------------------:|--------------------------:|
| ml.g5.48xlarge            | lmi             |                       57 |                            60 |                         8 |
| ml.p4d.24xlarge           | lmi             |                       26 |                           983 |                       256 |
| ml.g5.48xlarge            | lmi-accelerated |                       33 |                            43 |                         8 |
| ml.p4d.24xlarge (Default) | lmi-accelerated |                       10 |                           635 |                        64 |

-- With identity-based policy that allows the pricing:GetProducts action: No need to reboot the kernel, the output will be cached so future requests can use.

NOTE: After adding the policy, need to give it at least 10 seconds to propagate.

jumpstart_model.display_benchmark_metrics()

| Instance Type             | Config Name     |   Min Latency (ms/token) |   Max Throughput (tokens/sec) |   Max Concurrency (count) |   Instance Rate (USD/Hrs) |
|:--------------------------|:----------------|-------------------------:|------------------------------:|--------------------------:|--------------------------:|
| ml.g5.48xlarge            | lmi             |                       57 |                            60 |                         8 |                     20.36 |
| ml.p4d.24xlarge           | lmi             |                       26 |                           983 |                       256 |                     38.67 |
| ml.g5.48xlarge            | lmi-accelerated |                       33 |                            43 |                         8 |                     20.36 |
| ml.p4d.24xlarge (Default) | lmi-accelerated |                       10 |                           635 |                        64 |                     38.67 |

Merge Checklist

Put an x in the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your pull request.

General

  • [ x] I have read the CONTRIBUTING doc
  • [x ] I certify that the changes I am introducing will be backward compatible, and I have discussed concerns about this, if any, with the Python SDK team
  • [x ] I used the commit message format described in CONTRIBUTING
  • [ x] I have passed the region in to all S3 and STS clients that I've initialized as part of this change.
  • [x ] I have updated any necessary documentation, including READMEs and API docs (if appropriate)

Tests

  • [x ] I have added tests that prove my fix is effective or that my feature works (if appropriate)
  • [x ] I have added unit and/or integration tests as appropriate to ensure backward compatibility of the changes
  • [x ] I have checked that my tests are not configured for a specific region or account (if appropriate)
  • [x ] I have used unique_name_from_base to create resource names in integ tests (if appropriate)

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

@makungaj1 makungaj1 marked this pull request as ready for review May 1, 2024 20:08
@makungaj1 makungaj1 self-assigned this May 1, 2024
@makungaj1 makungaj1 merged commit c05e919 into benchmark-feature-fixes May 2, 2024
@makungaj1 makungaj1 deleted the benchmark-feature-fixes-tests branch May 2, 2024 17:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@qiyunz qiyunz qiyunz left review comments

Assignees

@makungaj1 makungaj1

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@makungaj1 @qiyunz