windows 11 搭建 TensorFlow GPU 开发环境【RTX 3060】：2 -- 基于WSL2 docker 方式的使用_3060 wsl2 gpu_shiter的博客

link管理
链接快照平台
输入网页链接，自动生成快照
标签化管理网页链接
相关文章推荐
温柔的泡面 · 关于天水一朝，为何宋朝被称为“天水一朝”？_ ...· 1 月前 ·
玩篮球的松树 · 同等学力申硕及如何读博士_西南政法大学· 6 月前 ·
活泼的扁豆 · 网站推广120种方法电子书免费下载· 7 月前 ·
爱热闹的熊猫 · 最高人民法院关于审理侵害知识产权民事案件适用 ...· 9 月前 ·
不敢表白的紫菜 · 摄影小组队名高级_头条· 1 年前 ·
简介
使用 wsl 的docker 进行深度学习与原生方式的对比
主要步骤
1.安装 wsl-2 版本的windows NVIDIA驱动
2. 在wsl-2 中安装 docker 及 NVIDIA 容器
测试1，simple container
测试2：Jupyter Notebooks
问题：为啥 jupyter notebook 的这个docker 调用巨慢无比？？？
目前我看官网主要推荐docker 方式了，那我们就用docker 方式试试。而且网上的安装教程也是docker 的居多【官方给出了一个教程】，我们也要与时俱进。
下面是我机器wsl kernel的版本： 可见是没有最新，只有更新哈！
season@season:~$ uname -r
5.10.16.3-microsoft-standard-WSL2
官方文档： 
https://docs.nvidia.com/cuda/wsl-user-guide/index.html 
使用 wsl 的docker 进行深度学习与 原生方式的对比
 
PyTorch MNIST 测试，这是一个有目的的小型玩具机器学习示例，它强调了保持 GPU 忙碌以达到满意的 WSL2性能的重要性。与原生 Linux 一样，工作负载越小，就越有可能由于启动 GPU 进程的开销而导致性能下降。这种退化在 WSL2上更为明显，并且与原生 Linux 的规模不同。
  
从图中可以看出如果batch size小的话，很多时间会消耗在CUDA调用上，batch size=8的时候，时间消耗会是native CUDA的138%。如果提高batch size，让CUDA充分忙碌，性能可以接近native！
 以下是原文链接。 
https://developer.nvidia.com/blog/leveling-up-cuda-performance-on-wsl2-with-new-enhancements/ 
1.安装 wsl-2 版本的windows NVIDIA驱动
 
cuda 驱动 on wsl
 https://developer.nvidia.com/cuda/wsl/download
 
 不知道为啥，这个没有特别说明和wsl 有啥关系。。。我已经有驱动了，这个不知道装了个啥。做如图所示的选择。 
特别注意，在wsl-2 中安装 cuda toolkit 要使用如下脚本：
 红框处是单独的选项 
wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-wsl-ubuntu.pin
sudo mv cuda-wsl-ubuntu.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/11.5.1/local_installers/cuda-repo-wsl-ubuntu-11-5-local_11.5.1-1_amd64.deb
sudo dpkg -i cuda-repo-wsl-ubuntu-11-5-local_11.5.1-1_amd64.deb
sudo apt-key add /var/cuda-repo-wsl-ubuntu-11-5-local/7fa2af80.pub
sudo apt-get update
sudo apt-get -y install cuda
安装好以后进行测试：Black-Scholes模型，简称B-S模型，是一种对金融产品估价的数学模型。 
cd /usr/local/cuda-11.5/samples/4_Finance/BlackScholes
season@season:/usr/local/cuda-11.5/samples/4_Finance/BlackScholes$ ll
total 60
drwxr-xr-x  4 root root  4096 Nov 24 23:07 ./
drwxr-xr-x 10 root root  4096 Nov 24 23:07 ../
drwxr-xr-x  2 root root  4096 Nov 24 23:07 .vscode/
-rw-r--r--  1 root root  8382 Sep 21 01:38 BlackScholes.cu
-rw-r--r--  1 root root  2787 Sep 21 01:38 BlackScholes_gold.cpp
-rw-r--r--  1 root root  3646 Sep 21 01:38 BlackScholes_kernel.cuh
-rw-r--r--  1 root root 13454 Sep 21 01:38 Makefile
-rw-r--r--  1 root root  1859 Sep 21 01:38 NsightEclipse.xml
drwxr-xr-x  2 root root  4096 Nov 24 23:07 doc/
-rw-r--r--  1 root root   189 Sep 21 01:38 readme.txt
season@season:/usr/local/cuda-11.5/samples/4_Finance/BlackScholes$ sudo make BlackScholes
>>> GCC Version is greater or equal to 5.1.0 <<<
/usr/local/cuda-11.5/bin/nvcc -ccbin g++ -I../../common/inc  -m64    -maxrregcount=16 --threads 0 --std=c++11 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_37,code=sm_37 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_52,code=sm_52 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_70,code=sm_70 -gencode arch=compute_75,code=sm_75 -gencode arch=compute_80,code=sm_80 -gencode arch=compute_86,code=sm_86 -gencode arch=compute_86,code=compute_86 -o BlackScholes.o -c BlackScholes.cu
nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).
ptxas warning : For profile sm_86 adjusting per thread register count of 16 to lower bound of 24
ptxas warning : For profile sm_75 adjusting per thread register count of 16 to lower bound of 24
ptxas warning : For profile sm_70 adjusting per thread register count of 16 to lower bound of 24
ptxas warning : For profile sm_80 adjusting per thread register count of 16 to lower bound of 24
/usr/local/cuda-11.5/bin/nvcc -ccbin g++ -I../../common/inc  -m64    -maxrregcount=16 --threads 0 --std=c++11 -gencode arch=compute_35,code=sm_35 -gencode arch=compute_37,code=sm_37 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_52,code=sm_52 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_70,code=sm_70 -gencode arch=compute_75,code=sm_75 -gencode arch=compute_80,code=sm_80 -gencode arch=compute_86,code=sm_86 -gencode arch=compute_86,code=compute_86 -o BlackScholes_gold.o -c BlackScholes_gold.cpp
nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).
/usr/local/cuda-11.5/bin/nvcc -ccbin g++   -m64      -gencode arch=compute_35,code=sm_35 -gencode arch=compute_37,code=sm_37 -gencode arch=compute_50,code=sm_50 -gencode arch=compute_52,code=sm_52 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_70,code=sm_70 -gencode arch=compute_75,code=sm_75 -gencode arch=compute_80,code=sm_80 -gencode arch=compute_86,code=sm_86 -gencode arch=compute_86,code=compute_86 -o BlackScholes BlackScholes.o BlackScholes_gold.o
nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).
mkdir -p ../../bin/x86_64/linux/release
cp BlackScholes ../../bin/x86_64/linux/release
season@season:/usr/local/cuda-11.5/samples/4_Finance/BlackScholes$ ./BlackScholes
[./BlackScholes] - Starting...
GPU Device 0: "Ampere" with compute capability 8.6
Initializing data...
...allocating CPU memory for options.
...allocating GPU memory for options.
...generating input data in CPU mem.
...copying input data to GPU mem.
Data init done.
Executing Black-Scholes GPU kernel (512 iterations)...
Options count             : 8000000
BlackScholesGPU() time    : 0.261508 msec
Effective memory bandwidth: 305.918207 GB/s
Gigaoptions per second    : 30.591821
BlackScholes, Throughput = 30.5918 GOptions/s, Time = 0.00026 s, Size = 8000000 options, NumDevsUsed = 1, Workgroup = 128
Reading back GPU results...
Checking the results...
...running CPU calculations.
Comparing the results...
L1 norm: 1.741792E-07
Max absolute error: 1.192093E-05
Shutting down...
...releasing GPU memory.
...releasing CPU memory.
Shutdown done.
[BlackScholes] - Test Summary
NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.
Test passed
2. 在wsl-2 中安装 docker 及 NVIDIA 容器
 
安装标准docker 
curl https://get.docker.com | sh
之后输出： 
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100




    
 18617  100 18617    0     0  25294      0 --:--:-- --:--:-- --:--:-- 25294
# Executing docker install script, commit: 93d2499759296ac1f9c510605fef85052a2c32be
WSL DETECTED: We recommend using Docker Desktop for Windows.
Please get Docker Desktop from https://www.docker.com/products/docker-desktop
You may press Ctrl+C now to abort this script.
+ sleep 20
+ sudo -E sh -c apt-get update -qq >/dev/null
+ sudo -E sh -c DEBIAN_FRONTEND=noninteractive apt-get install -y -qq apt-transport-https ca-certificates curl >/dev/null
+ sudo -E sh -c curl -fsSL "https://download.docker.com/linux/ubuntu/gpg" | gpg --dearmor --yes -o /usr/share/keyrings/docker-archive-keyring.gpg
+ sudo -E sh -c echo "deb [arch=amd64 signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu focal stable" > /etc/apt/sources.list.d/docker.list
+ sudo -E sh -c apt-get update -qq >/dev/null
+ sudo -E sh -c DEBIAN_FRONTEND=noninteractive apt-get install -y -qq --no-install-recommends  docker-ce-cli docker-scan-plugin docker-ce >/dev/null
+ version_gte 20.10
+ [ -z  ]
+ return 0
+ sudo -E sh -c DEBIAN_FRONTEND=noninteractive apt-get install -y -qq docker-ce-rootless-extras >/dev/null
================================================================================
To run Docker as a non-privileged user, consider setting up the
Docker daemon in rootless mode for your user:
    dockerd-rootless-setuptool.sh install
Visit https://docs.docker.com/go/rootless/ to learn about rootless mode.
To run the Docker daemon as a fully privileged service, but granting non-root
users access, refer to https://docs.docker.com/go/daemon-access/
WARNING: Access to the remote API on a privileged Docker daemon is equivalent
         to root access on the host. Refer to the 'Docker daemon attack surface'
         documentation for details: https://docs.docker.com/go/attack-surface/
================================================================================
装完之后发现，这个突然出现的玩意也是挺占用内存的
  
安装 NVIDIA Container 以及 Toolkit 
$ distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
$ curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
$ curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
$ curl -s -L https://nvidia.github.io/libnvidia-container/experimental/$distribution/libnvidia-container-experimental.list | sudo tee /etc/apt/sources.list.d/libnvidia-container-experimental.list
$ sudo apt-get update
$ sudo apt-get install -y nvidia-docker2
下面我们继续安装官网给出的测试docker 进行一下测试。
 https://docs.nvidia.com/cuda/wsl-user-guide/index.html 
测试1，simple container
 
season@season:~$ sudo service docker stop
[sudo] password for season:
 * Docker already stopped - file /var/run/docker-ssd.pid not found.
season@season:~$ sudo service docker start
 * Starting Docker: docker                                                                                       [ OK ]
season@season:~$ sudo docker run --gpus all nvcr.io/nvidia/k8s/cuda-sample:nbody nbody -gpu -benchmark
Unable to find image 'nvcr.io/nvidia/k8s/cuda-sample:nbody' locally
nbody: Pulling from nvidia/k8s/cuda-sample
d519e2592276: Pull complete
d22d2dfcfa9c: Pull complete
b3afe92c540b: Pull complete
b25f8d7adb24: Pull complete
ddb025f124b9: Pull complete
fe72fda9c19e: Pull complete
c6a265e4ffa3: Pull complete
c931a9542ebf: Pull complete
f7eb321dd245: Pull complete
d67fd954fbd5: Pull complete
Digest: sha256:a2117f5b8eb3012076448968fd1790c6b63975c6b094a8bd51411dee0c08440d
Status: Downloaded newer image for nvcr.io/nvidia/k8s/cuda-sample:nbody
Run "nbody -benchmark [-numbodies=<numBodies>]" to measure performance.
        -fullscreen       (run n-body simulation in fullscreen mode)
        -fp64             (use double precision floating point values for simulation)
        -hostmem          (stores simulation data in host memory)
        -benchmark        (run benchmark to measure performance)
        -numbodies=<N>    (number of bodies (>= 1) to run in simulation)
        -device=<d>       (where d=0,1,2.... for the CUDA device to use)
        -numdevices=<i>   (where i=(number of CUDA devices > 0) to use for simulation)
        -compare          (compares simulation results running once on the default GPU and once on the CPU)
        -cpu              (run n-body simulation on the CPU)
        -tipsy=<file.bin> (load a tipsy model file for simulation)
NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.
> Windowed mode
> Simulation data stored in video memory
> Single precision floating point simulation
> 1 Devices used for simulation
GPU Device 0: "Ampere" with compute capability 8.6
> Compute 8.6 CUDA device: [NVIDIA GeForce RTX 3060 Laptop GPU]
30720 bodies, total time for 10 iterations: 25.952 ms
= 363.636 billion interactions per second
= 7272.727 single-precision GFLOP/s at 20 flops per interaction
season@season:~$
测试2：Jupyter Notebooks
 
启动脚本： 
docker run -it --gpus all -p 8888:8888 tensorflow/tensorflow:latest-gpu-py3-jupyter
测试效果，感觉跑结果交互起来非常缓慢，下面这段程序大概跑了5-10分钟 
import tensorflow as tf
version = tf.__version__
gpu_ok = tf.test.is_gpu_available()
WARNING:tensorflow:From <ipython-input-3-11c8f8f7a7c6>:1: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.config.list_physical_devices('GPU')` instead.
tf.config.list_physical_devices('GPU')
[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]
 print("tf version:",version,"\nuse GPU",gpu_ok)
tf version: 2.1.0 
use GPU True
tf.reduce_sum(tf.random.normal([1000, 1000]))
<tf.Tensor: shape=(), dtype=float32, numpy=1321.2869>
问题：为啥 jupyter notebook 的这个docker 调用巨慢无比？？？
 
姑且把日志贴出来，大佬分析分析啥原因，会不是时间时区不一样？造成反映慢。。。




    
 
[I 16:29:45.604 NotebookApp] 302 GET /?token=99b4290f4d03e9fe537a7fe14872d20f2a50f8a23c81c247 (172.17.0.1) 0.55ms
[I 16:29:49.983 NotebookApp] Creating new notebook in
[I 16:29:50.024 NotebookApp] Writing notebook-signing key to /root/.local/share/jupyter/notebook_secret
[I 16:29:50.496 NotebookApp] Kernel started: 24d0d681-bfac-4723-9e73-541c5c4f2443
2021-11-25 16:30:05.794652: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libnvinfer.so.6
2021-11-25 16:30:05.824203: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libnvinfer_plugin.so.6
2021-11-25 16:30:29.686851: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 AVX512F FMA
2021-11-25 16:30:29.700683: I tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency: 2303995000 Hz
2021-11-25 16:30:29.702117: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x484f3f0 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2021-11-25 16:30:29.702147: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version
2021-11-25 16:30:29.704570: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcuda.so.1
2021-11-25 16:30:30.076279: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:30:30.076510: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x48c1ac0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2021-11-25 16:30:30.076562: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): NVIDIA GeForce RTX 3060 Laptop GPU, Compute Capability 8.6
2021-11-25 16:30:30.077105: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:30:30.077137: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1555] Found device 0 with properties:
pciBusID: 0000:01:00.0 name: NVIDIA GeForce RTX 3060 Laptop GPU computeCapability: 8.6
coreClock: 1.702GHz coreCount: 30 deviceMemorySize: 6.00GiB deviceMemoryBandwidth: 312.97GiB/s
2021-11-25 16:30:30.077168: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
2021-11-25 16:30:30.077183: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10
2021-11-25 16:30:30.111039: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcufft.so.10
2021-11-25 16:30:30.116390: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcurand.so.10
2021-11-25 16:30:30.162423: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusolver.so.10
2021-11-25 16:30:30.167253: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusparse.so.10
2021-11-25 16:30:30.167342: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudnn.so.7
2021-11-25 16:30:30.167733: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:30:30.167962: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:30:30.168006: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1697] Adding visible gpu devices: 0
2021-11-25 16:30:30.168473: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
[I 16:31:50.494 NotebookApp] Saving file at /Untitled.ipynb
2021-11-25 16:33:20.133916: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1096] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-11-25 16:33:20.133958: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1102]      0
2021-11-25 16:33:20.133980: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] 0:   N
2021-11-25 16:33:20.135139: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:33:20.135166: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1324] Could not identify NUMA node of platform GPU id 0, defaulting to 0.  Your kernel may not have been built with NUMA support.
2021-11-25 16:33:20.135374: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:33:20.135442: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1241] Created TensorFlow device (/device:GPU:0 with 4788 MB memory) -> physical GPU (device: 0, name: NVIDIA GeForce RTX 3060 Laptop GPU, pci bus id: 0000:01:00.0, compute capability: 8.6)
2021-11-25 16:33:20.152265: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:33:20.152312: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1555] Found device 0 with properties:
pciBusID: 0000:01:00.0 name: NVIDIA GeForce RTX 3060 Laptop GPU computeCapability: 8.6
coreClock: 1.702GHz coreCount: 30 deviceMemorySize: 6.00GiB deviceMemoryBandwidth: 312.97GiB/s
2021-11-25 16:33:20.152357: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
2021-11-25 16:33:20.152377: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10
2021-11-25 16:33:20.152406: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcufft.so.10
2021-11-25 16:33:20.152427: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcurand.so.10
2021-11-25 16:33:20.152446: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusolver.so.10
2021-11-25 16:33:20.152484: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusparse.so.10
2021-11-25 16:33:20.152507: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudnn.so.7
2021-11-25 16:33:20.152726: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:33:20.152927: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:33:20.152951: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1697] Adding visible gpu devices: 0
[I 16:33:51.283 NotebookApp] Saving file at /Untitled.ipynb
2021-11-25 16:35:09.708118: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:35:09.708166: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1555] Found device 0 with properties:
pciBusID: 0000:01:00.0 name: NVIDIA GeForce RTX 3060 Laptop GPU computeCapability: 8.6
coreClock: 1.702GHz coreCount: 30 deviceMemorySize: 6.00GiB deviceMemoryBandwidth: 312.97GiB/s
2021-11-25 16:35:09.708200: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
2021-11-25 16:35:09.708219: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10
2021-11-25 16:35:09.708245: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcufft.so.10
2021-11-25 16:35:09.708253: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcurand.so.10
2021-11-25 16:35:09.708289: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusolveeerer.eeeeer.so.10
2021-11-25 16:35:09.708310: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusparse.so.10
2021-11-25 16:35:09.708317: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudnn.so.7
2021-11-25 16:35:09.708564: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:35:09.708926: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:35:09.708955: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1697] Adding visible gpu devices: 0
2021-11-25 16:35:09.709019: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1096] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-11-25 16:35:09.709023: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1102]      0
2021-11-25 16:35:09.709027: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] 0:   N
2021-11-25 16:35:09.709380: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:35:09.709410: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1324] Could not identify NUMA node of platform GPU id 0, defaulting to 0.  Your kernel may not have been built with NUMA support.
2021-11-25 16:35:09.709641: E tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:967] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node
Your kernel may have been built without NUMA support.
2021-11-25 16:35:09.709682: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1241] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 4788 MB memory) -> physical GPU (device: 0, name: NVIDIA GeForce RTX 3060 Laptop GPU, pci bus id: 0000:01:00.0, compute capability: 8.6)
安装WSL2，官方文档说的比较清楚了 
https://docs.microsoft.com/zh-cn/windows/wsl/install 
5步搭建wsl2+cuda+docker解决windows深度学习开发问题 
https://zhuanlan.zhihu.com/p/408403790 
Windows+WSL2+CUDA+Docker 
https://blog.csdn.net/fleaxin/article/details/108911522 
tensor flow 官方gpu 支持文档 
https://tensorflow.google.cn/install/gpu 
cuda 官方指导 
https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/ 
本文主要参考：CUDA on WSL User Guide 
https://docs.nvidia.com/cuda/wsl-user-guide/index.html 
11.5 版本的文档有docker 说明 
https://docs.nvidia.com/cuda/archive/11.5.0/wsl-user-guide/index.html#ch05-running-containers
启用 WSL2
dism.exe /online /enable-feature /featurename:Microsoft-Windows-Subsystem-Linux /all /norestart
dism.exe /online /enable-feature /featurename:VirtualMachineP
                                    本流程在Windows 11下进行，也适用于最新版本的Windows 10。基于Windows操作系统，对于日常使用和同时使用多个其他的开发环境更加友好，上手难度低；而基于Linux的深度学习，生态成熟且案例也更多，因此WSL2是一个非常不错的选择。同时本文兼顾了基于Cuda的深度学习开发。
                                    在《物体检测快速入门系列(2)-Windows部署GPU深度学习开发环境》一文中已经描述了如何在Windows环境下部署GPU深度学习开发环境，但是要我推荐的话，我还是倾向于docker环境，无需安装cuda、cuDNN，docker镜像安装完毕后，就都好了，一键部署好之后，可以随意迁移，再也不用环境发愁了............................
                                    目录背景问题解决方法
已经配置好了wsl2和docker desktop，因C盘空间不够卸载了Ubuntu20.04，后来又重新安装了Ubuntu20.04（nVidia-docker需要）
可以在powershell下使用docker，但是不能在Ubuntu下运行，提示如下
$ docker
	Command 'docker' not found, but can be installed with:
	sudo apt install docker.io
$ sudo apt ins
https://docs.nvidia.com/cuda/wsl-user-guide/index.html#installing-wip
1.wsl2安装
wsl2安装实际上非常简单，大体上来说可以分为两步
安装激活wsl2功能
其实这里可以直接参考官方的参考文档。但是其实这里我并没有直接使用命令就能成功，此处参考了另一个官方文档，成功安装了wsl2。
安装linux的发行版
linux发行版的安装在上面的官方教程里也有提到，其实就是直接到微软的应用商店下载安装对应的版本就行。（我这
                                    以前捣鼓过wsl，即Windows下的Linux子系统，但兼容性依然比不过原生的Linux系统，使用cmake等命令会出现奇怪的问题。最近听说wsl2出来了，而且也可以在wsl上安装nvidia显卡驱动了，有网友实测跑深度学习模型速度能比Windows的快一倍左右，哈哈这就必须得捣鼓捣鼓了，如果兼容性真的没问题的话，那可比虚拟机或双系统要爽多了~
                                    原因:can't create unix socket /var/run/docker.sock: is a directory/var/run/docker.sock是一个目录,导致docker启动失败解决过程:删除docker.sock目录,service docker start 启动docker服务启动过程好像出现了长时间的阻塞,后不得已强制结束,Ctrl+c查看/var/run/目录下已...
                                    基于docker在Ubuntu上搭建TensorFlow-GPU计算环境
由于实验室的服务器有多人共享使用，而不同人的代码对应的keras和tensorflow版本不一致，所以对应的cuda版本也不相同，因此，考虑使用docker安装自己的容器，这样就可以避免共享cuda版本不一致造成的麻烦。（不过有贴子说使用docker的话，GPU性能只能发挥80%，所以有利有弊吧）
安装docker
首先，检...
使用 wsl 的docker 进行深度学习与 原生方式的对比

1.安装 wsl-2 版本的windows NVIDIA驱动

2. 在wsl-2 中安装 docker 及 NVIDIA 容器

测试1，simple container

测试2：Jupyter Notebooks

问题：为啥 jupyter notebook 的这个docker 调用巨慢无比？？？

使用 wsl 的docker 进行深度学习与原生方式的对比
