次の方法で共有

【Azure】ジョブの実行が失敗するようになった【Machine Learning Studio】

TM107710 40 評価のポイント
2025-09-09T07:05:52.0866667+00:00

理由は不明ですが、ジョブを実行すると失敗で終了するようになりました。

設定としては、「自動ML」で、タスクは「時系列の予測」としております。

ジョブを以前の通り正常終了するための方法をご教示頂けないでしょうか。

【ジョブのエラーログ】

Execution failed. User process 'python' exited with status code 1.

Please check log file 'user_logs/std_log.txt' for error details. Error: Traceback (most recent call last):

File "setup_xxx0000001.py", line 26, in <module>

from azureml.train.automl import fit_pipeline

ImportError: cannot import name 'fit_pipeline' from 'azureml.train.automl' (/azureml-envs/azureml-automl/lib/python3.9/site-packages/azureml/train/automl/init.py)

【std_log.txtの内容】

Importing plotly failed. Interactive plots will not work.

2025-09-09:05:55:04,776 WARNING [_remote_script.py:17] Encountered exception when importing one or more remote wrappers.

v2 driver import failed with exception: cannot import name 'setup_wrapper' from 'azureml.train.automl._remote_script'

(/azureml-envs/azureml-automl/lib/python3.9/site-packages/azureml/train/automl/_remote_script.py). Falling back to v1 driver.

Traceback (most recent call last):

File "setup_xxx0000001.py", line 11, in <module>

from azureml.train.automl._remote_script import setup_wrapper

ImportError: cannot import name 'setup_wrapper' from 'azureml.train.automl._remote_script'

(/azureml-envs/azureml-automl/lib/python3.9/site-packages/azureml/train/automl/_remote_script.py)

Cleaning up all outstanding Run operations, waiting 300.0 seconds

1 items cleaning up...

Cleanup took 0.06326103210449219 seconds

Traceback (most recent call last):

File "setup_xxx0000001.py", line 11, in <module>

from azureml.train.automl._remote_script import setup_wrapper

ImportError: cannot import name 'setup_wrapper' from 'azureml.train.automl._remote_script'

(/azureml-envs/azureml-automl/lib/python3.9/site-packages/azureml/train/automl/_remote_script.py)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):

File "setup_xxx0000001.py", line 26, in <module>

from azureml.train.automl import fit_pipeline

ImportError: cannot import name 'fit_pipeline' from 'azureml.train.automl'

(/azureml-envs/azureml-automl/lib/python3.9/site-packages/azureml/train/automl/init.py)

Azure
Azure

Microsoft が管理する世界のデータ センター ネットワークを介してアプリケーションとサービスを構築、配置、および管理するインフラストラクチャおよびクラウド コンピューティング プラットフォーム。


1 件の回答

並べ替え方法: 最も役に立つ
  1. Pavankumar Purilla 11,575 評価のポイント Microsoft 外部スタッフ モデレーター
    2025-09-11T02:59:38.3433333+00:00

    Hi TM107710,
    詳細をご共有いただきありがとうございます。失敗した実行 (AzureML-ai-ml-automl:30) と成功した実行 (AzureML-ai-ml-automl:22) の違いは、確かに重要な要素です。これらの番号は、AutoML がジョブを実行するためにバックグラウンドで使用するランタイム環境 (Docker イメージ) を示しています。サービスは定期的にこれらの環境を更新し、新しい機能、ライブラリ、または修正プログラムを導入します。ただし、場合によっては、新しい環境に一時的な問題や非互換性があり、以前は同じデータと構成で動作していたにもかかわらず、ジョブが失敗することがあります。

    環境バージョン :22 で全く同じデータセットを使用してジョブが成功したという事実は、データセット自体が根本原因ではないことを強く示唆しています。むしろ、バージョン :30 での失敗は、その特定の環境ビルドにおける一時的な問題に関連していた可能性が高いです。これらの種類の問題は通常、イメージにパッチが適用されるかロールバックされるとサービス側で解決されるため、後で同じジョブを再試行すると正常に完了する理由が説明できます。

    AutoML環境のバージョンを手動で管理したり固定したりする必要はありません。サービスが自動的に適切なランタイムをプロビジョニングします。このような場合の推奨アプローチは、予期しない障害が発生した場合にジョブを再試行し、特定の環境バージョンで問題が継続的に発生するかどうかを監視することです。あるバージョンでのみ繰り返し障害が発生し、他のバージョンでは正常に動作する場合は、ジョブログをキャプチャしてサポートリクエストを送信してください。エンジニアリングチームが影響を受ける環境を調査し、修復します。

    この回答は役に立ちましたか?

    0 件のコメント コメントはありません

お客様の回答

質問作成者は回答に "承認済み"、モデレーターは "おすすめ" とマークできます。これにより、ユーザーは作成者の問題が回答によって解決したことを把握できます。