Skip to content

app-mlperf-inference-intel

Automatically generated README for this automation recipe: app-mlperf-inference-intel

Category: Modular MLPerf benchmarks

License: Apache 2.0

  • CM meta description for this script: _cm.yaml
  • Output cached? False

Reuse this script in your project

Install MLCommons CM automation meta-framework

Pull CM repository with this automation recipe (CM script)

cm pull repo mlcommons@cm4mlops

cmr "reproduce mlcommons mlperf inference harness intel-harness intel intel-harness intel" --help

Run this script

Run this script via CLI
cm run script --tags=reproduce,mlcommons,mlperf,inference,harness,intel-harness,intel,intel-harness,intel[,variations] [--input_flags]
Run this script via CLI (alternative)
cmr "reproduce mlcommons mlperf inference harness intel-harness intel intel-harness intel [variations]" [--input_flags]
Run this script from Python
import cmind

r = cmind.access({'action':'run'
              'automation':'script',
              'tags':'reproduce,mlcommons,mlperf,inference,harness,intel-harness,intel,intel-harness,intel'
              'out':'con',
              ...
              (other input keys for this script)
              ...
             })

if r['return']>0:
    print (r['error'])
Run this script via Docker (beta)
cm docker script "reproduce mlcommons mlperf inference harness intel-harness intel intel-harness intel[variations]" [--input_flags]

Variations

  • No group (any combination of variations can be selected)

    Click here to expand this section.

    • _bs.#
      • ENV variables:
        • ML_MLPERF_MODEL_BATCH_SIZE: #
    • _v3.1
      • ENV variables:
        • CM_MLPERF_INFERENCE_CODE_VERSION: v3.1
  • Group "device"

    Click here to expand this section.

    • _cpu (default)
      • ENV variables:
        • CM_MLPERF_DEVICE: cpu
  • Group "framework"

    Click here to expand this section.

    • _pytorch (default)
      • ENV variables:
        • CM_MLPERF_BACKEND: pytorch
        • CM_MLPERF_BACKEND_LIB_NAMESPEC: pytorch
  • Group "loadgen-batchsize"

    Click here to expand this section.

    • _batch_size.#
      • ENV variables:
        • CM_MLPERF_LOADGEN_BATCH_SIZE: #
  • Group "loadgen-scenario"

    Click here to expand this section.

    • _multistream
      • ENV variables:
        • CM_MLPERF_LOADGEN_SCENARIO: MultiStream
    • _offline
      • ENV variables:
        • CM_MLPERF_LOADGEN_SCENARIO: Offline
    • _server
      • ENV variables:
        • CM_MLPERF_LOADGEN_SCENARIO: Server
    • _singlestream
      • ENV variables:
        • CM_MLPERF_LOADGEN_SCENARIO: SingleStream
  • Group "model"

    Click here to expand this section.

    • _bert-99
      • ENV variables:
        • CM_MODEL: bert-99
        • CM_SQUAD_ACCURACY_DTYPE: float32
        • CM_NOT_ML_MODEL_STARTING_WEIGHTS_FILENAME: https://zenodo.org/record/3750364/files/bert_large_v1_1_fake_quant.onnx
    • _bert-99.9
      • ENV variables:
        • CM_MODEL: bert-99.9
        • CM_NOT_ML_MODEL_STARTING_WEIGHTS_FILENAME: https://zenodo.org/record/3733910/files/model.onnx
    • _gptj-99
      • ENV variables:
        • CM_MODEL: gptj-99
        • CM_NOT_ML_MODEL_STARTING_WEIGHTS_FILENAME: https://zenodo.org/record/3733910/files/model.onnx
        • CM_ML_MODEL_WEIGHTS_DATA_TYPE: int8
        • CM_ML_MODEL_INPUTS_DATA_TYPE: int8
    • _gptj-99.9
      • ENV variables:
        • CM_MODEL: gptj-99.9
        • CM_NOT_ML_MODEL_STARTING_WEIGHTS_FILENAME: https://zenodo.org/record/3733910/files/model.onnx
    • _resnet50 (default)
      • ENV variables:
        • CM_MODEL: resnet50
        • dataset_imagenet_preprocessed_input_square_side: 224
        • ml_model_has_background_class: YES
        • ml_model_image_height: 224
        • loadgen_buffer_size: 1024
        • loadgen_dataset_size: 50000
        • CM_BENCHMARK: STANDALONE_CLASSIFICATION
    • _retinanet
      • ENV variables:
        • CM_MODEL: retinanet
        • CM_ML_MODEL_STARTING_WEIGHTS_FILENAME: https://zenodo.org/record/6617981/files/resnext50_32x4d_fpn.pth
        • dataset_imagenet_preprocessed_input_square_side: 224
        • ml_model_image_height: 800
        • ml_model_image_width: 800
        • loadgen_buffer_size: 64
        • loadgen_dataset_size: 24576
        • CM_BENCHMARK: STANDALONE_OBJECT_DETECTION
  • Group "network-mode"

    Click here to expand this section.

    • _network-server
      • ENV variables:
        • CM_MLPERF_NETWORK_RUN_MODE: network-server
    • _standalone (default)
      • ENV variables:
        • CM_MLPERF_NETWORK_RUN_MODE: standalone
  • Group "network-run-mode"

    Click here to expand this section.

    • _network-client
      • ENV variables:
        • CM_MLPERF_NETWORK_RUN_MODE: network-client
  • Group "power-mode"

    Click here to expand this section.

    • _maxn
      • ENV variables:
        • CM_MLPERF_NVIDIA_HARNESS_MAXN: True
    • _maxq
      • ENV variables:
        • CM_MLPERF_NVIDIA_HARNESS_MAXQ: True
  • Group "precision"

    Click here to expand this section.

    • _fp32
      • ENV variables:
        • CM_IMAGENET_ACCURACY_DTYPE: float32
    • _int4
    • _uint8
  • Group "run-mode"

    Click here to expand this section.

    • _build-harness
      • ENV variables:
        • CM_LOCAL_MLPERF_INFERENCE_INTEL_RUN_MODE: build_harness
    • _calibration
      • ENV variables:
        • CM_LOCAL_MLPERF_INFERENCE_INTEL_RUN_MODE: calibration
    • _run-harness (default)
      • ENV variables:
        • CM_LOCAL_MLPERF_INFERENCE_INTEL_RUN_MODE: run_harness
  • Group "sut"

    Click here to expand this section.

    • _sapphire-rapids.112c
      • ENV variables:
        • WARMUP: --warmup
    • _sapphire-rapids.24c
  • Group "version"

    Click here to expand this section.

    • _v4.0 (default)
      • ENV variables:
        • CM_MLPERF_INFERENCE_CODE_VERSION: v4.0
Default variations

_cpu,_pytorch,_resnet50,_run-harness,_standalone,_v4.0

Script flags mapped to environment

  • --count=valueCM_MLPERF_LOADGEN_QUERY_COUNT=value
  • --max_batchsize=valueCM_MLPERF_LOADGEN_MAX_BATCHSIZE=value
  • --mlperf_conf=valueCM_MLPERF_CONF=value
  • --mode=valueCM_MLPERF_LOADGEN_MODE=value
  • --multistream_target_latency=valueCM_MLPERF_LOADGEN_MULTISTREAM_TARGET_LATENCY=value
  • --offline_target_qps=valueCM_MLPERF_LOADGEN_OFFLINE_TARGET_QPS=value
  • --output_dir=valueCM_MLPERF_OUTPUT_DIR=value
  • --performance_sample_count=valueCM_MLPERF_LOADGEN_PERFORMANCE_SAMPLE_COUNT=value
  • --rerun=valueCM_RERUN=value
  • --scenario=valueCM_MLPERF_LOADGEN_SCENARIO=value
  • --server_target_qps=valueCM_MLPERF_LOADGEN_SERVER_TARGET_QPS=value
  • --singlestream_target_latency=valueCM_MLPERF_LOADGEN_SINGLESTREAM_TARGET_LATENCY=value
  • --skip_preprocess=valueCM_SKIP_PREPROCESS_DATASET=value
  • --skip_preprocessing=valueCM_SKIP_PREPROCESS_DATASET=value
  • --target_latency=valueCM_MLPERF_LOADGEN_TARGET_LATENCY=value
  • --target_qps=valueCM_MLPERF_LOADGEN_TARGET_QPS=value
  • --user_conf=valueCM_MLPERF_USER_CONF=value

Default environment

These keys can be updated via --env.KEY=VALUE or env dictionary in @input.json or using script flags.

  • CM_BATCH_COUNT: 1
  • CM_BATCH_SIZE: 1
  • CM_FAST_COMPILATION: yes
  • CM_MLPERF_LOADGEN_SCENARIO: Offline
  • CM_MLPERF_LOADGEN_MODE: performance
  • CM_SKIP_PREPROCESS_DATASET: no
  • CM_SKIP_MODEL_DOWNLOAD: no
  • CM_MLPERF_SUT_NAME_IMPLEMENTATION_PREFIX: intel
  • CM_MLPERF_SKIP_RUN: no
  • verbosity: 1
  • loadgen_trigger_cold_run: 0

Native script being run


Script output

cmr "reproduce mlcommons mlperf inference harness intel-harness intel intel-harness intel [variations]" [--input_flags] -j