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
 
本文同步分享在 博客“shiter”（CSDN）。
如有侵权，请联系 support@oschina.cn 删除。
本文参与“OSC源创计划”，欢迎正在阅读的你也加入，一起分享。
                                        展开阅读全文 
                            看了 wangEditor 的公告，如鲠在喉。去年七月，我在一篇《关于剔除 layedit 组件》的公告中，还推荐了几款 editor 组件用于替代，其中就包括了 wangEditor, 转眼之间，仿若时空交错，不免有些感慨。 在国内由个人发起的开源项目，似乎都很难跳出相同的宿命，不断有人走进这个赛道，但能抵达终点并完成突破的却屈指可数。Layui 曾经同样倒在了赛道，2021 年宣布关站之前，Layui 的百度指数还一度领先 Bootstrap, 如此一个拥有广泛用户群体的 UI 库，本该迎来新的突破，却在疾跑中戛然止步，至今令人迷惑。人们看到的，是一篇充满悲情色彩的公告，而公告的背后，是创作者在面对内外交织的困境中不得已做出的举措，当我们没有足够的力量冲破眼前的障碍，除了停下来避开它，还有别的更优选择么。你很难想象除此之外还有多少历史包袱… 譬如，也是由于种种原因，当初 Layui 在 Github 和 Gitee 待处理的 Issue 数量，不下于 2000，各种议题参次不齐，我差不多花了半年时间去逐一审阅，多少个日夜消耗，多少次自我情绪的对抗… 就不多提了。 尽管这两年来，Layui 的受众者已呈断崖式流失，但正因为小众，反而如释重负，甚至让我重新找回了一些开源的纯粹。 共勉 🙂
                            我作为该项目的开发团队成员之一，在此说明: 1. 这个项目是完全由自己编写的，没有借助商业的帮助，没有他人冒充，也不存在父母帮我们编写后给我们挂名的问题。同时因为我们都是中学生，没有多少资金，也不可能进行"买广告"等不正当的商业行为; 2. 我们自己做的项目绝对没有博眼球，单纯就是想对Windows在网页上进行模仿和创新; 3. 我们的实力并不是那么差，我们开发团队成员曾获得C++二级全市第一、CSP-J国赛一等奖、蓝桥杯省赛一等奖等多个奖项，成员每人编程能力全国前4%-10%，并且我们人均精通前端和后端，就拿我来举例，我能熟练使用git命令行，从克隆仓库，提交、推送更改到拉取合并、暂存、设置远程、配置ssh密钥都烂熟于心，熟悉Linux系统，从查看当前目录下文件和文件夹到使用命令行手动安装驱动都熟悉，熟悉NodeJS，会下载、导入模块和程序编写与运行，精通HTML, CSS和JS，基本的东西都能灵活运用，并且熟悉Python, C, C++和汇编语言; 4. 我们项目链接在 https://github.com/tjy-gitnub/win12，现已经获得超1500颗Stars
                            本人原创作者，在此解释声明一二： 1. 我的父母还没有无耻到做这种无耻的事情，我也对我的能力有信心，目前也不需要这些偷鸡摸狗的见不得光的东西来造假，请各位不要以小人之心度君子之腹。 2. 我今年初三，项目是从初一开始做的，各位不相信的欢迎上github看提交记录，可以看一下初版和现在的区别，UI和js都有不小更改。 3. 团队组成：今年一位初三，一位初一，一位5+4制的初一。更新记录中明确记录了哪些功能是哪位所开发的。 4. 我在github上，bilibili上没有受到过任何一个人的质疑，我对中国的网络环境表示蔑视。 5. 本项目的初中只是为了兴趣，没有想到火了起来。 我不理解你们是如何通过代码读出作者年龄的？难道是语文考试要加入代码阅读赏析的题目了吗？真心觉得很厉害。 6. 不愿与某些人同流，也无众位深厚阅历经验，只望能得清白之名。感谢大家让我懂得了何乃人情世故，孰谓世态炎凉，世俗红尘。你们给我的人生上了重要的一课。人间哪有什么真善美啊，呵。社会的病胎罢了。
                            用过`php`, 有几个缺点用不下去： - `php`的动态语言特性造成经常`IDE`不能很好的跳转源码位置。 - 语言本身不支持调试，需要用插件，上`Apache`。 比如我想测试一个方法。 - 对某个文件进行单元测试麻烦。 - 变量命名很乱。 动态语言团队开发就是噩梦，比如突然从某个变量里访问某个`key`,都不知道这个`key`哪里来的，`IDE`支持不好的话得全部搜索一遍。 - 想要运行时进入第三方库并调试，难。 - 性能如果纯`php`并不高，有谁会经常写`C`扩展？ - 如果想用线程，比如非实时定时任务，邮件发送，数据统计等。本身标准库不支持线程。 - 除了做网站方便点其他程序开发难做。比如想做个分析`Excel`的命令行小工具，不能打包为`exe`并发给别人用。 都是学语言，其他语言比如`java`,`go`还能做非网站程序，有的选学PHP干嘛？当然目前网站部署成本`php`最低，也支持热部署这个目前静态语言还超越不了。如果要学动态语言，用`javascript`开发网站后端也不错啊，还能把经验用在开发前端网页上。