【Python】Anaconda软件的conda.exe基本使用

1.为什么还需要 Anaconda？有以下3个原因：

1）Anaconda 附带了一大批常用数据科学包，它附带了 conda、Python 和 150 多个科学包及其依赖项。因此你可以立即开始处理数据。

2）管理包

Anaconda 是在 conda（一个包管理器和环境管理器）上发展出来的。

在数据分析中，你会用到很多第三方的包，而conda（包管理器）可以很好的帮助你在计算机上安装和管理这些包，包括安装、卸载和更新包。

3）管理环境

为什么需要管理环境呢？

比如你在A项目中用了 Python 2，而新的项目B老大要求使用Python 3，而同时安装两个Python版本可能会造成许多混乱和错误。这时候 conda就可以帮助你为不同的项目建立不同的运行环境。

还有很多项目使用的包版本不同，比如不同的pandas版本，不可能同时安装两个 Numpy 版本，你要做的应该是，为每个 Numpy 版本创建一个环境，然后项目的对应环境中工作。这时候conda就可以帮你做到。

(https://www.zhihu.com/question/58033789/answer/254673663)

 

2.Anaconda的安装教程：

https://blog.csdn.net/qq_38628350/article/details/79045640

https://blog.csdn.net/qq_36015370/article/details/79484455/

https://www.zhihu.com/question/58033789

后面再补充百度云资源

anaconda官网：https://www.anaconda.com/download/

 

3.conda的基本使用：

conda.exe在“D:\Users\Leon_PC\Anaconda3\Scripts”下可以找到，

打开windows控制台（win+r，输入cmd，回车）

如果之前没有配置好环境，那么定位到“D:\Users\Leon_PC\Anaconda3\Scripts”

 cd /d D:\Users\Leon_PC\Anaconda3\Scripts

 

（1）conda下载的操作，关于下载速度的问题

• 查看conda的下载源

conda config --show channels

• 从channel中安装包时显示channel的url，这样就可以知道包的安装来源了

conda config --set show_channel_urls yes

• 添加清华下载源

conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/

• 删除某个下载源

conda config --remove channels

比如：

conda config --remove channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/

一些不太能用但是网上基本都有的源：

$ conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
$ conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
##  Conda Forge
$ conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
## msys2
$ conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
## bioconda
$ conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/
## menpo
$ conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/menpo/

 

 

（2）conda创建虚拟环境等系列管理环境的操作

• 查看存在哪些虚拟环境

conda env list 或 conda info -e 

• 创建新的虚拟环境：tensorflow_gpu是虚拟环境的名称，可以任意设置，其中 -n参数表示环境的名称

 conda create -n tensorflow_gpu python=3.6

查看conda create有哪些option参数

conda create -h

usage: conda-script.py create [-h] [--clone ENV] [-n ENVIRONMENT | -p PATH]
                              [-c CHANNEL] [--use-local] [--override-channels]
                              [--repodata-fn REPODATA_FNS]
                              [--strict-channel-priority]
                              [--no-channel-priority]
                              [--no-deps | --only-deps] [--no-pin] [--copy]
                              [--no-shortcuts] [-C] [-k] [--offline] [-d]
                              [--json] [-q] [-v] [-y] [--download-only]
                              [--show-channel-urls] [--file FILE]
                              [--no-default-packages] [--dev]
                              [package_spec [package_spec ...]]

Create a new conda environment from a list of specified packages. To use the created environment, use 'conda activate envname' look in that directory first.  This command requires either the -n NAME or -p PREFIX option.

Options:

positional arguments:
  package_spec          Packages to install or update in the conda
                        environment.

optional arguments:
  -h, --help            Show this help message and exit.
  --clone ENV           Path to (or name of) existing local environment.
  --file FILE           Read package versions from the given file. Repeated
                        file specifications can be passed (e.g. --file=file1
                        --file=file2).
  --dev                 Use `sys.executable -m conda` in wrapper scripts
                        instead of CONDA_EXE. This is mainly for use during
                        tests where we test new conda source against old
                        Python versions.

Target Environment Specification:
  -n ENVIRONMENT, --name ENVIRONMENT
                        Name of environment.
  -p PATH, --prefix PATH
                        Full path to environment location (i.e. prefix).

Channel Customization:
  -c CHANNEL, --channel CHANNEL
                        Additional channel to search for packages. These are
                        URLs searched in the order they are given (including
                        local directories using the 'file://' syntax or simply
                        a path like '/home/conda/mychan' or '../mychan').
                        Then, the defaults or channels from .condarc are
                        searched (unless --override-channels is given). You
                        can use 'defaults' to get the default packages for
                        conda. You can also use any name and the .condarc
                        channel_alias value will be prepended. The default
                        channel_alias is http://conda.anaconda.org/.
  --use-local           Use locally built packages. Identical to '-c local'.
  --override-channels   Do not search default or .condarc channels. Requires
                        --channel.
  --repodata-fn REPODATA_FNS
                        Specify name of repodata on remote server. Conda will
                        try whatever you specify, but will ultimately fall
                        back to repodata.json if your specs are not
                        satisfiable with what you specify here. This is used
                        to employ repodata that is reduced in time scope. You
                        may pass this flag more than once. Leftmost entries
                        are tried first, and the fallback to repodata.json is
                        added for you automatically.

Solver Mode Modifiers:
  --strict-channel-priority
                        Packages in lower priority channels are not considered
                        if a package with the same name appears in a higher
                        priority channel.
  --no-channel-priority
                        Package version takes precedence over channel
                        priority. Overrides the value given by `conda config
                        --show channel_priority`.
  --no-deps             Do not install, update, remove, or change
                        dependencies. This WILL lead to broken environments
                        and inconsistent behavior. Use at your own risk.
  --only-deps           Only install dependencies.
  --no-pin              Ignore pinned file.
  --no-default-packages
                        Ignore create_default_packages in the .condarc file.

Package Linking and Install-time Options:
  --copy                Install all packages using copies instead of hard- or
                        soft-linking.
  --no-shortcuts        Don't install start menu shortcuts

Networking Options:
  -C, --use-index-cache
                        Use cache of channel index files, even if it has
                        expired.
  -k, --insecure        Allow conda to perform "insecure" SSL connections and
                        transfers. Equivalent to setting 'ssl_verify' to
                        'false'.
  --offline             Offline mode. Don't connect to the Internet.

Output, Prompt, and Flow Control Options:
  -d, --dry-run         Only display what would have been done.
  --json                Report all output as json. Suitable for using conda
                        programmatically.
  -q, --quiet           Do not display progress bar.
  -v, --verbose         Can be used multiple times. Once for INFO, twice for
                        DEBUG, three times for TRACE.
  -y, --yes             Do not ask for confirmation.
  --download-only       Solve an environment and ensure package caches are
                        populated, but exit prior to unlinking and linking
                        packages into the prefix.
  --show-channel-urls   Show channel urls. Overrides the value given by `conda
                        config --show show_channel_urls`.

Examples:

    conda create -n myenv sqlite

 

• 激活环境（进入到虚拟环境中）

activate tensorflow_gpu

• 退出虚拟环境

conda deactivate

• 删除某个虚拟环境

conda remove -n tensorflow_gpu --all

conda remove -h
usage: conda-script.py remove [-h] [-n ENVIRONMENT | -p PATH] [-c CHANNEL]
                              [--use-local] [--override-channels]
                              [--repodata-fn REPODATA_FNS] [--all]
                              [--features] [--force-remove] [--no-pin] [-C]
                              [-k] [--offline] [-d] [--json] [-q] [-v] [-y]
                              [--dev]
                              [package_name [package_name ...]]

Remove a list of packages from a specified conda environment.

    This command will also remove any package that depends on any of the
    specified packages as well---unless a replacement can be found without
    that dependency. If you wish to skip this dependency checking and remove
    just the requested packages, add the '--force' option. Note however that
    this may result in a broken environment, so use this with caution.

Options:

positional arguments:
  package_name          Package names to remove from the environment.

optional arguments:
  -h, --help            Show this help message and exit.
  --dev                 Use `sys.executable -m conda` in wrapper scripts
                        instead of CONDA_EXE. This is mainly for use during
                        tests where we test new conda source against old
                        Python versions.

Target Environment Specification:
  -n ENVIRONMENT, --name ENVIRONMENT
                        Name of environment.
  -p PATH, --prefix PATH
                        Full path to environment location (i.e. prefix).

Channel Customization:
  -c CHANNEL, --channel CHANNEL
                        Additional channel to search for packages. These are
                        URLs searched in the order they are given (including
                        local directories using the 'file://' syntax or simply
                        a path like '/home/conda/mychan' or '../mychan').
                        Then, the defaults or channels from .condarc are
                        searched (unless --override-channels is given). You
                        can use 'defaults' to get the default packages for
                        conda. You can also use any name and the .condarc
                        channel_alias value will be prepended. The default
                        channel_alias is http://conda.anaconda.org/.
  --use-local           Use locally built packages. Identical to '-c local'.
  --override-channels   Do not search default or .condarc channels. Requires
                        --channel.
  --repodata-fn REPODATA_FNS
                        Specify name of repodata on remote server. Conda will
                        try whatever you specify, but will ultimately fall
                        back to repodata.json if your specs are not
                        satisfiable with what you specify here. This is used
                        to employ repodata that is reduced in time scope. You
                        may pass this flag more than once. Leftmost entries
                        are tried first, and the fallback to repodata.json is
                        added for you automatically.

Solver Mode Modifiers:
  --all                 Remove all packages, i.e., the entire environment.
  --features            Remove features (instead of packages).
  --force-remove, --force
                        Forces removal of a package without removing packages
                        that depend on it. Using this option will usually
                        leave your environment in a broken and inconsistent
                        state.
  --no-pin              Ignore pinned file.

Networking Options:
  -C, --use-index-cache
                        Use cache of channel index files, even if it has
                        expired.
  -k, --insecure        Allow conda to perform "insecure" SSL connections and
                        transfers. Equivalent to setting 'ssl_verify' to
                        'false'.
  --offline             Offline mode. Don't connect to the Internet.

Output, Prompt, and Flow Control Options:
  -d, --dry-run         Only display what would have been done.
  --json                Report all output as json. Suitable for using conda
                        programmatically.
  -q, --quiet           Do not display progress bar.
  -v, --verbose         Can be used multiple times. Once for INFO, twice for
                        DEBUG, three times for TRACE.
  -y, --yes             Do not ask for confirmation.

Examples:

    conda remove -n myenv scipy

 

• 虚拟环境重命名

conda 其实没有重命名指令，实现重命名是通过 clone 完成的，分两步：

1. 先 clone 一份 new name 的环境
2. 删除 old name 的环境

conda create -n new_env --clone tensorflow_gpu

conda remove -n tensorflow_gpu --all

 

• 查看虚拟环境中通过conda install安装了哪些包

conda list

conda list -h
usage: conda-script.py list [-h] [-n ENVIRONMENT | -p PATH] [--json] [-v] [-q]
                            [--show-channel-urls] [-c] [-f] [--explicit]
                            [--md5] [-e] [-r] [--no-pip]
                            [regex]

List linked packages in a conda environment.

Options:

positional arguments:
  regex                 List only packages matching this regular expression.

optional arguments:
  -h, --help            Show this help message and exit.
  --show-channel-urls   Show channel urls. Overrides the value given by `conda
                        config --show show_channel_urls`.
  -c, --canonical       Output canonical names of packages only. Implies --no-
                        pip.
  -f, --full-name       Only search for full names, i.e., ^$.
  --explicit            List explicitly all installed conda packaged with URL
                        (output may be used by conda create --file).
  --md5                 Add MD5 hashsum when using --explicit
  -e, --export          Output requirement string only (output may be used by
                        conda create --file).
  -r, --revisions       List the revision history and exit.
  --no-pip              Do not include pip-only installed packages.

Target Environment Specification:
  -n ENVIRONMENT, --name ENVIRONMENT
                        Name of environment.
  -p PATH, --prefix PATH
                        Full path to environment location (i.e. prefix).

Output, Prompt, and Flow Control Options:
  --json                Report all output as json. Suitable for using conda
                        programmatically.
  -v, --verbose         Use once for info, twice for debug, three times for
                        trace.
  -q, --quiet           Do not display progress bar.

Examples:

List all packages in the current environment:

    conda list

List all packages installed into the environment 'myenv':

    conda list -n myenv

Save packages for future use:

    conda list --export > package-list.txt

Reinstall packages from an export file:

    conda create -n myenv --file package-list.txt

 

• 查看虚拟环境中通过pip install安装了哪些包

pip list

• 查看虚拟环境中的python版本（activate env先进入虚拟环境）

python --version

 

• 查找conda源中存在的安装包

 conda search tensorflow

conda search -h
usage: conda-script.py search [-h] [--envs] [-i] [--subdir SUBDIR]
                              [-c CHANNEL] [--use-local] [--override-channels]
                              [--repodata-fn REPODATA_FNS] [-C] [-k]
                              [--offline] [--json] [-v] [-q]

Search for packages and display associated information.
    The input is a MatchSpec, a query language for conda packages.
    See examples below.

Options:

optional arguments:
  -h, --help            Show this help message and exit.
  --envs                Search all of the current user's environments. If run
                        as Administrator (on Windows) or UID 0 (on unix),
                        search all known environments on the system.
  -i, --info            Provide detailed information about each package.
  --subdir SUBDIR, --platform SUBDIR
                        Search the given subdir. Should be formatted like
                        'osx-64', 'linux-32', 'win-64', and so on. The default
                        is to search the current platform.

Channel Customization:
  -c CHANNEL, --channel CHANNEL
                        Additional channel to search for packages. These are
                        URLs searched in the order they are given (including
                        local directories using the 'file://' syntax or simply
                        a path like '/home/conda/mychan' or '../mychan').
                        Then, the defaults or channels from .condarc are
                        searched (unless --override-channels is given). You
                        can use 'defaults' to get the default packages for
                        conda. You can also use any name and the .condarc
                        channel_alias value will be prepended. The default
                        channel_alias is http://conda.anaconda.org/.
  --use-local           Use locally built packages. Identical to '-c local'.
  --override-channels   Do not search default or .condarc channels. Requires
                        --channel.
  --repodata-fn REPODATA_FNS
                        Specify name of repodata on remote server. Conda will
                        try whatever you specify, but will ultimately fall
                        back to repodata.json if your specs are not
                        satisfiable with what you specify here. This is used
                        to employ repodata that is reduced in time scope. You
                        may pass this flag more than once. Leftmost entries
                        are tried first, and the fallback to repodata.json is
                        added for you automatically.

Networking Options:
  -C, --use-index-cache
                        Use cache of channel index files, even if it has
                        expired.
  -k, --insecure        Allow conda to perform "insecure" SSL connections and
                        transfers. Equivalent to setting 'ssl_verify' to
                        'false'.
  --offline             Offline mode. Don't connect to the Internet.

Output, Prompt, and Flow Control Options:
  --json                Report all output as json. Suitable for using conda
                        programmatically.
  -v, --verbose         Use once for info, twice for debug, three times for
                        trace.
  -q, --quiet           Do not display progress bar.

Examples:

Search for a specific package named 'scikit-learn':

    conda search scikit-learn

Search for packages containing 'scikit' in the package name:

    conda search *scikit*

Note that your shell may expand '*' before handing the command over to conda.
Therefore it is sometimes necessary to use single or double quotes around the query.

    conda search '*scikit'
    conda search "*scikit*"

Search for packages for 64-bit Linux (by default, packages for your current
platform are shown):

    conda search numpy[subdir=linux-64]

Search for a specific version of a package:

    conda search 'numpy>=1.12'

Search for a package on a specific channel

    conda search conda-forge::numpy
    conda search 'numpy[channel=conda-forge, subdir=osx-64]'

• 查找当前电脑中所有虚拟环境中的某个包，比如tensorflow（在本机内查找包）

conda search tensorflow --envs

列出所有虚拟环境中关于tensorflow的安装包信息

 

• 更新包

conda update xxx （更新某一个包）

conda update --all（更新所有包）

 

• 卸载包

conda remove 或 conda uninstall

 

（3）新建完虚拟环境之后安装spyder

• 激活虚拟环境

activate tensorflow_gpu

• 安装spyder

conda install spyder

会安装一个专门配对虚拟环境的spyder，这样你就可以在spyder里面import 虚拟环境中的tensorflow，而不需要用到pycharm去设置python源什么的，当然那样也是可以的，但是我比较常用spyder作为IDE

 

（4）安装tensorflow要注意的问题

• 通过conda install不要用pip，不然可能不能正常import

conda install tensorflow-gpu==1.13.1

conda 安装将自动安装 GPU 支持所需的 CUDA 和 CuDNN 库。pip 安装则需要手动安装这些库。人人喜欢一步到位，尤其是在下载与安装库这方面。

使用 pip 安装 TensorFlow 时，GPU 支持所需的 CUDA 和 CuDNN 库必须单独手动安装，增加了大量负担。而使用 conda 安装 GPU 加速版本的 TensorFlow 时，只需使用命令 conda install tensorflow-gpu，这些库就会自动安装成功，且版本与 tensorflow-gpu 包兼容。此外，conda 安装这些库的位置不会与通过其他方法安装的库的其他实例产生冲突。不管使用 pip 还是 conda 安装 GPU 支持的 TensorFlow，NVIDIA 驱动程序都必须单独安装。

对于 TensorFlow 的多个版本，conda 包可使用多种 CUDA 版本。例如，对于 TensorFlow 1.10.0 版本，conda 包支持可用的 CUDA 8.0、9.0 和 9.2 库。而 pip 包仅支持 CUDA 9.0 库。在不支持 CUDA 库最新版本的系统上运行时，这非常重要。最后，由于这些库是通过 conda 自动安装的，用户可轻松创建多个环境，并对比不同 CUDA 版本的性能。

（https://blog.csdn.net/abc13526222160/article/details/84673109）

 

（5）安装opencv要注意的问题

• 要用pip安装opencv-python，然后用conda安装opencv

https://blog.csdn.net/lyq_12/article/details/84783220

pip install opencv-python

conda install opencv