TensorFlow 的 JupyterLab 环境

TensorFlow 准备 JupyterLab 交互式笔记本环境，方便我们边写代码、边做笔记。

基础环境

以下是本文的基础环境，不详述安装过程了。

Ubuntu

• Ubuntu 18.04.5 LTS (Bionic Beaver)
  • ubuntu-18.04.5-desktop-amd64.iso

CUDA

• CUDA 11.2.2
  • cuda_11.2.2_460.32.03_linux.run
• cuDNN 8.1.1
  • libcudnn8_8.1.1.33-1+cuda11.2_amd64.deb
  • libcudnn8-dev_8.1.1.33-1+cuda11.2_amd64.deb
  • libcudnn8-samples_8.1.1.33-1+cuda11.2_amd64.deb

Anaconda

• Anaconda Python 3.8
  • Anaconda3-2020.11-Linux-x86_64.sh

1conda activate base

安装 JupyterLab

Anaconda 环境里已有，如下查看版本：

1jupyter --version

不然，如下进行安装：

1conda install -c conda-forge jupyterlab

安装 TensorFlow

创建虚拟环境 tf，再 pip 安装 TensorFlow：

1# create virtual environment
2conda create -n tf python=3.8 -y
3conda activate tf
4
5# install tensorflow
6pip install --upgrade pip
7pip install tensorflow

测试：

1$ python - <<EOF
2import tensorflow as tf
3print(tf.__version__, tf.test.is_built_with_gpu_support())
4print(tf.config.list_physical_devices('GPU'))
5EOF

12021-04-01 11:18:17.719061: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0
22.4.1 True
32021-04-01 11:18:18.437590: I tensorflow/compiler/jit/xla_cpu_device.cc:41] Not creating XLA devices, tf_xla_enable_xla_devices not set
42021-04-01 11:18:18.437998: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcuda.so.1
52021-04-01 11:18:18.458471: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
62021-04-01 11:18:18.458996: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties:
7pciBusID: 0000:01:00.0 name: GeForce RTX 2060 computeCapability: 7.5
8coreClock: 1.35GHz coreCount: 30 deviceMemorySize: 5.79GiB deviceMemoryBandwidth: 245.91GiB/s
92021-04-01 11:18:18.459034: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.11.0
102021-04-01 11:18:18.461332: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.11
112021-04-01 11:18:18.461362: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.11
122021-04-01 11:18:18.462072: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10
132021-04-01 11:18:18.462200: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10
142021-04-01 11:18:18.462745: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10
152021-04-01 11:18:18.463241: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.11
162021-04-01 11:18:18.463353: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.8
172021-04-01 11:18:18.463415: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
182021-04-01 11:18:18.463854: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:941] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
192021-04-01 11:18:18.464170: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0
20[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

Solution: Could not load dynamic library 'libcusolver.so.10'

1cd /usr/local/cuda/lib64
2sudo ln -sf libcusolver.so.11 libcusolver.so.10

安装 IPython kernel

在虚拟环境 tf 里，安装 ipykernel 与 Jupyter 交互。

1# install ipykernel (conda new environment)
2conda activate tf
3conda install ipykernel -y
4python -m ipykernel install --user --name tf --display-name "Python TF"
5
6# run JupyterLab (conda base environment with JupyterLab)
7conda activate base
8jupyter lab

另一种方式，可用 nb_conda 扩展，其于笔记里会激活 Conda 环境：

1# install ipykernel (conda new environment)
2conda activate tf
3conda install ipykernel -y
4
5# install nb_conda (conda base environment with JupyterLab)
6conda activate base
7conda install nb_conda -y
8# run JupyterLab
9jupyter lab

最后，访问 http://localhost:8888/ ：

参考

• Install TensorFlow 2
  • Build from source
  • GPU support
• Install TensorFlow - Anaconda
  • anaconda / packages / tensorflow
• Installing the IPython kernel

GoCoding 个人实践的经验分享，可关注公众号！