共用方式為


SynapseSparkStep 類別

注意

這是實驗性類別,可以隨時變更。 請參閱 https://aka.ms/azuremlexperimental 以取得詳細資訊。

建立提交和執行 Python 腳本的 Azure ML Synapse 步驟。

建立在 synapse Spark 集區上執行 Spark 作業的 Azure ML 管線步驟。

繼承
azureml.pipeline.core._synapse_spark_step_base._SynapseSparkStepBase
SynapseSparkStep

建構函式

SynapseSparkStep(file, source_directory, compute_target, driver_memory, driver_cores, executor_memory, executor_cores, num_executors, name=None, app_name=None, environment=None, arguments=None, inputs=None, outputs=None, conf=None, py_files=None, jars=None, files=None, allow_reuse=True, version=None)

參數

名稱 Description
file
必要
str

與source_directory相對的 synapse 腳本名稱。

source_directory
必要
str

資料夾,其中包含步驟中使用的 Python 腳本、conda env 和其他資源。

compute_target
必要

要使用的計算目標。

driver_memory
必要
str

用於驅動程式進程的記憶體數量。

driver_cores
必要
int

要用於驅動程式進程的核心數目。

executor_memory
必要
str

每個執行程式進程要使用的記憶體數量。

executor_cores
必要
int

每個執行程式要使用的核心數目。

num_executors
必要
int

要為此會話啟動的執行程式數目。

name
必要
str

步驟的名稱。 如果未指定, file 則會使用 。

app_name
必要
str

用來提交 Spark 作業的應用程式名稱。

environment
必要

稍後版本將支援 AML 環境。

arguments
必要

Synapse 腳本檔案的命令列引數。

inputs
必要

輸入的清單。

outputs
必要

輸出清單。

conf
必要

Spark 組態屬性。

py_files
必要

要在此會話中使用的 Python 檔案,livy API 的參數。

files
必要

要在此會話中使用的檔案,livy API 的參數。

allow_reuse
必要

指出當使用相同的設定重新執行時,步驟是否應該重複使用先前的結果。

version
必要
str

選擇性版本戳記,表示步驟的功能變更。

file
必要
str

相對於 的 Synapse 腳本 source_directory 名稱。

source_directory
必要
str

資料夾,其中包含步驟中使用的 Python 腳本、conda env 和其他資源。

compute_target
必要

要使用的計算目標。

driver_memory
必要
str

用於驅動程式進程的記憶體數量。

driver_cores
必要
int

要用於驅動程式進程的核心數目。

executor_memory
必要
str

每個執行程式進程要使用的記憶體數量。

executor_cores
必要
int

每個執行程式要使用的核心數目。

num_executors
必要
int

要為此會話啟動的執行程式數目。

name
必要
str

步驟的名稱。 如果未指定, file 則會使用 。

app_name
必要
str

用來提交 Apache Spark 作業的應用程式名稱。

environment
必要

將在此 SynapseSparkStep 中運用的 AML 環境。

arguments
必要

Synapse 腳本檔案的命令列引數。

inputs
必要

輸入的清單。

outputs
必要

輸出清單。

conf
必要

Spark 組態屬性。

py_files
必要

要在此會話中使用的 Python 檔案,livy API 的參數。

jars
必要

要用於此會話的 Jar 檔案,livy API 的參數。

files
必要

要在此會話中使用的檔案,livy API 的參數。

allow_reuse
必要

指出當使用相同的設定重新執行時,步驟是否應該重複使用先前的結果。

version
必要
str

選擇性版本戳記,表示步驟的功能變更。

備註

SynapseSparkStep 是在 Synapse Spark 集區上執行 Python Spark 作業的基本內建步驟。 它會採用主要檔案名和其他選擇性參數,例如腳本、計算目標、輸入和輸出的引數。

使用 SynapseSparkStep 的最佳做法是針對腳本和與步驟相關聯的任何相依檔案使用不同的資料夾,並使用 參數指定該資料夾 source_directory 。 遵循此最佳做法有兩個優點。 首先,它有助於減少為步驟建立的快照集大小,因為只會快照步驟所需的快照集。 第二,如果沒有任何變更會觸發重新上傳快照集的 , source_directory 則可以重複使用上一次執行的步驟輸出。


   from azureml.core import Dataset
   from azureml.pipeline.steps import SynapseSparkStep
   from azureml.data import HDFSOutputDatasetConfig

   # get input dataset
   input_ds = Dataset.get_by_name(workspace, "weather_ds").as_named_input("weather_ds")

   # register pipeline output as dataset
   output_ds = HDFSOutputDatasetConfig("synapse_step_output",
                                       destination=(ws.datastores['datastore'],"dir")
                                       ).register_on_complete(name="registered_dataset")

   step_1 = SynapseSparkStep(
       name = "synapse_step",
       file = "pyspark_job.py",
       source_directory="./script",
       inputs=[input_ds],
       outputs=[output_ds],
       compute_target = "synapse",
       driver_memory = "7g",
       driver_cores = 4,
       executor_memory = "7g",
       executor_cores = 2,
       num_executors = 1,
       conf = {})

SynapseSparkStep 僅支援 DatasetConsumptionConfig 作為輸入,而 HDFSOutputDatasetConfig 做為輸出。

方法

create_node

建立 Synapse 腳本步驟的節點。

這個方法不適合直接使用。 使用此步驟具現化管線時,Azure ML 會自動傳遞透過此方法所需的參數,以便將步驟新增至代表工作流程的管線圖形。

create_node

建立 Synapse 腳本步驟的節點。

這個方法不適合直接使用。 使用此步驟具現化管線時,Azure ML 會自動傳遞透過此方法所需的參數,以便將步驟新增至代表工作流程的管線圖形。

create_node(graph, default_datastore, context)

參數

名稱 Description
graph
必要

要加入節點的繪圖物件。

default_datastore
必要

預設資料存放區。

context
必要
<xref:azureml.pipeline.core._GraphContext>

圖形內容。

傳回

類型 Description

已建立的節點。