深度学习tensorflow报错: Error polling for event status: failed to query event: CUDA_ERROR_ILLEGAL_ADDRESS

人工智能78

服务器环境:

最近在跑point cloud classification的demo PointASNL的时候,当batch_size设置比较大的时候,就会在训练中间出现以下错误:

2020-06-12 00:14:01.824110: E tensorflow/stream_executor/cuda/cuda_event.cc:29] Error polling for event status: failed to query event: CUDA_ERROR_ILLEGAL_ADDRESS: an illegal memory access was encountered
2020-06-12 00:14:01.824142: F tensorflow/core/common_runtime/gpu/gpu_event_mgr.cc:273] Unexpected Event status: 1

一开始以为是gpu的编程代码出现问题,反复检查之后发现并无错误。

从网上搜集资料之后隐约意识到应该是环境版本的问题。

在把cudnn 7.4.5降到cudnn 7.3.1之后,这个问题似乎解决了,希望别再出幺蛾子了。

Original: https://blog.csdn.net/qq_35692267/article/details/118420440
Author: 曼巴精神程序员
Title: 深度学习tensorflow报错: Error polling for event status: failed to query event: CUDA_ERROR_ILLEGAL_ADDRESS