tensorflow-gpu conda环境不工作于ubuntu-20.04

Question

使用conda在ubuntu20.04上安装tensorflow的说明如下：使用Anaconda安装Tensorflow

conda create --name tf tensorflow-gpu

成功创建了conda环境。但是，使用gpu的tensorflow无法工作：

 !18
python3 image_classification_efficientnet_fine_tuning.py
2021-04-25 13:08:58.235320: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory
2021-04-25 13:08:58.235341: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
Traceback (most recent call last):
  File "image_classification_efficientnet_fine_tuning.py", line 137, in <module>
    import tensorflow as tf
  File "/home/steve/.local/lib/python3.8/site-packages/tensorflow/__init__.py", line 436, in <module>
    _ll.load_library(_main_dir)
  File "/home/steve/.local/lib/python3.8/site-packages/tensorflow/python/framework/load_library.py", line 153, in load_library
    py_tf.TF_LoadLibrary(lib)
tensorflow.python.framework.errors_impl.NotFoundError: /home/steve/anaconda3/envs/tf/lib/python3.8/site-packages/tensorflow/core/kernels/libtfkernel_sobol_op.so: undefined symbol: _ZNK10tensorflow8OpKernel11TraceStringB5cxx11ERKNS_15OpKernelContextEb

让Tensorflow用cuda安装在ubuntu-20.04上是非常困难的。

Answer 1

在为virtual_environment创建TensorFlow之后，您需要在该虚拟环境中安装TensorFlow。

conda create -n tf-gpu tensorflow-gpu   # to create a virtual environment
conda activate tf-gpu                   # to activate the virtual environment

conda install tensorflow-gpu            # to install tensorflow-gpu

注意:每当您想运行'tf-gpu'代码时，请每次在anaconda中选择这个'tf-gpu'虚拟环境。

在执行上述代码后再试运行您的代码，如果问题仍然存在，请通知我们。