pipeline.yml

$schema: https://azuremlschemas.azureedge.net/latest/pipelineJob.schema.json
type: pipeline

description: Pipeline using AutoML Image Multiclass Classification task

display_name: pipeline-with-image-classification
experiment_name: pipeline-with-automl

settings:
  default_compute: azureml:gpu-cluster

inputs:
  image_multiclass_classification_training_data:
    type: mltable
    # Update the path, if prepare_data.py is using data_path other than "./data"
    path: data/training-mltable-folder
  image_multiclass_classification_validation_data:
    type: mltable
    # Update the path, if prepare_data.py is using data_path other than "./data"
    path: data/validation-mltable-folder

jobs:
  image_multiclass_classification_node:
    type: automl
    task: image_classification
    log_verbosity: info
    primary_metric: accuracy
    limits:
      timeout_minutes: 180
      max_trials: 10
      max_concurrent_trials: 2
    target_column_name: label
    training_data: ${{parent.inputs.image_multiclass_classification_training_data}}
    validation_data: ${{parent.inputs.image_multiclass_classification_validation_data}}
    sweep:
      sampling_algorithm: random
      early_termination:
        type: bandit
        evaluation_interval: 2
        slack_factor: 0.2
        delay_evaluation: 6
    search_space:
      - model_name:
          type: choice
          values: [vitb16r224, vits16r224]
        learning_rate:
          type: uniform
          min_value: 0.001
          max_value: 0.01
        number_of_epochs:
          type: choice
          values: [15, 30]

      - model_name:
          type: choice
          values: [seresnext, resnet50]
        learning_rate:
          type: uniform
          min_value: 0.001
          max_value: 0.01
        layers_to_freeze:
          type: choice
          values: [0, 2]
    training_parameters:
      early_stopping: True
      evaluation_frequency: 1
    # currently need to specify outputs "mlflow_model" explicitly to reference it in following nodes
    outputs:
      best_model:
        type: mlflow_model
  register_model_node:
    type: command
    component: file:./components/component_register_model.yaml
    inputs:
      model_input_path: ${{parent.jobs.image_multiclass_classification_node.outputs.best_model}}
      model_base_name: fridge_items_multiclass_classification_model