Jupyter笔记本和Anaconda-Navigator中内核数量的差异

Question

I am running python with Ananconda, and when I load my jupyter notebook, I see that there are 6 kernels available (see image):

Last login: Sun Sep 17 12:42:58 on ttys001
MacBook:~ user1$ jupyter notebook
[I 14:53:55.642 NotebookApp] [nb_conda_kernels] enabled, 6 kernels found
[I 14:53:56.470 NotebookApp] JupyterLab alpha preview extension loaded from /Users/user/anaconda/lib/python2.7/site-packages/jupyterlab
[I 14:53:56.471 NotebookApp] Running the core application with no additional extensions or settings
[I 14:53:57.374 NotebookApp] [nb_anacondacloud] enabled
[I 14:53:57.379 NotebookApp] [nb_conda] enabled
[I 14:53:57.451 NotebookApp] ✓ nbpresent HTML export ENABLED
[W 14:53:57.452 NotebookApp] ✗ nbpresent PDF export DISABLED: No module named nbbrowserpdf.exporters.pdf
[I 14:53:57.457 NotebookApp] Serving notebooks from local directory: /Users/user

But conda and anaconda-navigator show three environments:

conda info --envs
# conda environments:
#
P34                      /Users/user/anaconda/envs/P34
R                        /Users/user/anaconda/envs/R
root                  *  /Users/user/anaconda

Furthermore,

conda info --json

Returns this:

{
  "GID": 20, 
  "UID": 503, 
  "channels": [
    "https://conda.anaconda.org/anaconda-fusion/osx-64", 
    "https://conda.anaconda.org/anaconda-fusion/noarch", 
    "https://conda.anaconda.org/r/osx-64", 
    "https://conda.anaconda.org/r/noarch", 
    "https://repo.continuum.io/pkgs/free/osx-64", 
    "https://repo.continuum.io/pkgs/free/noarch", 
    "https://repo.continuum.io/pkgs/r/osx-64", 
    "https://repo.continuum.io/pkgs/r/noarch", 
    "https://repo.continuum.io/pkgs/pro/osx-64", 
    "https://repo.continuum.io/pkgs/pro/noarch"
  ], 
  "conda_build_version": "2.0.2", 
  "conda_env_version": "4.3.25", 
  "conda_location": "/Users/user/anaconda/lib/python2.7/site-packages/conda", 
  "conda_prefix": "/Users/user/anaconda", 
  "conda_private": false, 
  "conda_version": "4.3.25", 
  "default_prefix": "/Users/user/anaconda", 
  "env_vars": {
    "CIO_TEST": "<not set>", 
    "CONDA_DEFAULT_ENV": "<not set>", 
    "CONDA_ENVS_PATH": "<not set>", 
    "DYLD_LIBRARY_PATH": "<not set>", 
    "PATH": "/Users/user/Dropbox (Personal)/firefoxdriver_osx/bin:/Users/user/Dropbox (Personal)/chromedriver_osx/bin:/Users/user/anaconda/bin:/Library/Frameworks/Python.framework/Versions/Current/bin:/usr/local/bin:/usr/bin:/bin:/usr/sbin:/sbin:/opt/X11/bin:/Library/TeX/texbin", 
    "PYTHONHOME": "<not set>", 
    "PYTHONPATH": "<not set>"
  }, 
  "envs": [
    "/Users/user/anaconda/envs/P34", 
    "/Users/user/anaconda/envs/R"
  ], 
  "envs_dirs": [
    "/Users/user/anaconda/envs", 
    "/Users/user/.conda/envs"
  ], 
  "netrc_file": null, 
  "offline": false, 
  "pkgs_dirs": [
    "/Users/user/anaconda/pkgs", 
    "/Users/user/.conda/pkgs"
  ], 
  "platform": "osx-64", 
  "python_version": "2.7.13.final.0", 
  "rc_path": "/Users/user/.condarc", 
  "requests_version": "2.14.2", 
  "root_prefix": "/Users/user/anaconda", 
  "root_writable": true, 
  "site_dirs": [], 
  "sys.executable": "/Users/user/anaconda/bin/python", 
  "sys.prefix": "/Users/user/anaconda", 
  "sys.version": "2.7.13 |Anaconda 4.4.0 (x86_64)| (default, Dec 20 2016, 23:05:08) \n[GCC 4.2.1 Compatible Apple LLVM 6.0 (clang-600.0.57)]", 
  "sys_rc_path": "/Users/user/anaconda/.condarc", 
  "user_agent": "conda/4.3.25 requests/2.14.2 CPython/2.7.13 Darwin/16.7.0 OSX/10.12.6", 
  "user_rc_path": "/Users/user/.condarc"
}

How can I reconcile what the jupyter notebook has and what conda reports? How can I delete some of the environments reported by Jupyter notebook? This issue is creating with some libraries that I think are loaded in an environment, but end up in one of the "ghost" environments.

Answer 1

Conda ships with nb_conda_kernels , which bypass the normal kernel mechanism; thus not all program of your computer can see all the kernels.

nb_conda_kernels is meant to automatically expose any conda env as a possible kernel. It does (or at least did) that only for the notebook server, thus you seeing more kernels from within the notebook UI.

This has advantages : no more manual registering of kernels – and disadvantages : many other piece of software can't see al the kernels; notably, Atom, Nteract, Nbconvert, and other lower level tools as well, and the issues you are mentioning.

You can deactivate nb_conda_kernel by looking at your jupyter configuration files and remove the options anaconda has enabled; uninstall nb_conda_kernels. Then install kernelspec the classic way .

Answer 2

IPython kernels != conda environments. You might have environments with multiple kernels (as in your case environments 'P34' and 'R' have 2 kernels each - for R and Python), or there might be environments without IPython kernels at all.

If you want to remove a conda environment, you do that with conda env remove -n ENV_NAME command.

If you want to detach the kernel from Jupyter without removing the whole environment, you can delete the corresponding kernel spec folder. See my answer to your other question for the details on how to find the kernel spec folder.