components automl_hts_inference - Azure/azureml-assets GitHub Wiki
Enables inference for hts components.
Version: 0.0.7
View in Studio: https://ml.azure.com/registries/azureml/components/automl_hts_inference/version/0.0.7
Name | Description | Type | Default | Optional | Enum |
---|---|---|---|---|---|
raw_data | Folder URI with inference data. | uri_folder | |||
compute_name | Compute name for inference pipeline. | string | |||
max_nodes | Number of nodes in a compute cluster we will run the inference step on. | integer | |||
max_concurrency_per_node | Number of processes that will be run concurrently on any given node. This number should not be larger than 1/2 of the number of cores in an individual node in the specified cluster | integer | |||
parallel_step_timeout_in_seconds | The PRS step time out settings is seconds. | integer | 3600 | ||
train_run_id | The train run id used for training models that will be used to generate forecasts. | string | True | ||
train_experiment_name | The train experiment used for inference. | string | True | ||
forecast_mode | The forecast mode used for inference. The possible values are recursive and rolling . |
string | recursive | True | ['recursive', 'rolling'] |
forecast_step | The forecast step used for rolling forecast. See more details here: https://learn.microsoft.com/en-us/azure/machine-learning/how-to-auto-train-forecast?view=azureml-api-2#evaluating-model-accuracy-with-a-rolling-forecast | integer | 1 | True | |
allocation_method | Method that allocates forecasts within a hierarchy. Possible values are average_historical_proportions and proportions_of_historical_average | string | average_historical_proportions | ['average_historical_proportions', 'proportions_of_historical_average'] | |
forecast_level | Specify the level of the hierarchy for which you are interested in obtaining the forecast for. | string | |||
optional_train_metadata | Metadata from training run. | uri_folder | True | ||
forecast_quantiles | Space separated list of quantiles to get forecasts for forecast quantiles for forecasting jobs. It is applicable only when the forecast_mode is recursive. | string | True |
Name | Description | Type |
---|---|---|
run_output | Folder URI representing the location of the output data | uri_folder |
evaluation_configs | The evaluation configs. | uri_file |
evaluation_data | The evaluation data. | uri_file |