logo
Loading...

unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize - Cupoy

錯誤訊息:UnknownError: 2 root error(s) found.  (0) Unk...

unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize

2020/05/31 上午 02:24
機器學習共學討論版
Nick
觀看數:25
回答數:1
收藏數:0

錯誤訊息:

UnknownError: 2 root error(s) found.
  (0) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
     [[{{node conv2d_4/convolution}}]]
     [[Mean_1/_213]]
  (1) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
     [[{{node conv2d_4/convolution}}]]
0 successful operations.
0 derived errors ignored.

硬體資訊:

  • docker images:

        nvidia/cuda           latest                             9e47e9dfcb9a        5 months ago        2.83GB

  • nvidia-smi
Sat May 30 17:52:39 2020       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.82       Driver Version: 440.82       CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.| Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 960M    Off  | 00000000:01:00.0 Off |                  N/|
| N/A   45C    P5    N//  N/|   2004MiB /  2004MiB |     17%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+
  • OS Platform and Distribution : Ubuntu 16.4
  • TensorFlow installed from (source or binary): Docker hub
  • TensorFlow version: 1.15-gpu
  • Python version: 3.6.8
  • Installed using virtualenv? pip? conda?: pip
  • Bazel version (if compiling from source): 0.18
  • GCC/Compiler version (if compiling from source): gcc 5.4.0
  • CUDA version:  CUDA- 10.2
  • GPU model and memory: GeForce GTX960 major: 4 minor: 1 memoryClockRate(GHz): 1.8225 2GB


安裝方法:

1.參考官網,使用docker安裝(https://www.tensorflow.org/install/docker)


自我檢測:

1. 使用 tf.test.is_gpu_available(cuda_only=True, min_cuda_compute_capability=None) 檢測過

  得到True


想請問一下,關於這個錯誤有那幾種可能,還有哪些解決方案

回答列表

  • 2020/06/01 上午 00:13
    張維元 (WeiYuan)
    贊同數:0
    不贊同數:0
    留言數:1

    嗨,這個有可能是 tensorflow 版本的問題,建議確認一下版本。


    ```

    import tensorflow as tf 

    tf.__version__

    ```


    如果這個回答對你有幫助請主動點選「有幫助」的按鈕,也可以追蹤我的GITHUB帳號。若還有問題的話,也歡迎繼續再追問或者把你理解的部分整理上來,我都會提供你 Review 和 Feedback 😃😃😃