Azure Synapse中PyTorch导入引发Spark会话失败求助

环境与依赖库 #

我的Azure Synapse Notebook使用以下Python库：

import pandas as pd
import numpy as np
from sqlalchemy import create_engine, text
import sqlalchemy as sa
from azure.core.credentials import AzureKeyCredential
from azure.ai.textanalytics import TextAnalyticsClient
from sentence_transformers import SentenceTransformer, util
import time
import torch

from notebookutils import mssparkutils

问题现象 #

本月起，执行完上述导入代码的单元格后，会话停止并弹出如下警告：

/home/trusted-service-user/cluster-env/env/lib/python3.10/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: libc10_cuda.so: cannot open shared object file: No such file or directory
  warn(f"Failed to load image Python extension: {e}")
2024-11-19 11:31:17.000433: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  SSE4.1 SSE4.2 AVX AVX2 AVX512F FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.

随后Spark会话失败，提示Session failed. Run the notebook to start a new session.。该Notebook在2-3周前运行正常，此前即使出现上述警告也不影响使用。

已尝试的解决方案 #

• 回退Torch版本
• 强制使用CPU运行
  以上操作均未解决问题。

附加Spark驱动stderr日志 #

WARN TokenLibrary [pool-43-thread-2]: Access token cache miss or expired
2024-11-19 11:24:16,395 ERROR TokenLibrary [pool-43-thread-2]: Unable to determine host value from URI = tokenservice2.westeurope.azuresynapse.net:443. Using localhost as header value

WARN SQLConf [spark-listener-group-shared]: The SQL config 'spark.sql.legacy.replaceDatabricksSparkAvro.enabled' has been deprecated in Spark v3.2 and may be removed in the future. Use `.format("avro")` in `DataFrameWriter` or `DataFrameReader` instead.

2024-11-19 11:23:32,622 WARN AzureBlobFileSystemStore [Thread-32]: checkDnsEntry: blabla.dfs.core.windows.net not found in the file /etc/hosts.

请求协助排查并解决该会话失败问题。

内容的提问来源于stack exchange，提问作者Dimitar Petrov