A Jupyter kernel base class in Python which includes core magic functions (including help, command and file path completion, parallel and distributed processing, downloads, and much more).
See Jupyter's docs on wrapper kernels.
Additional magics can be installed within the new kernel package under a magics subpackage.
... and many others.
You can install Metakernel through pip:
pip install metakernel --upgrade
Installing metakernel from the conda-forge channel can be achieved by adding conda-forge to your channels with:
conda config --add channels conda-forge
Once the conda-forge channel has been enabled, metakernel can be installed with:
It is possible to list all of the versions of metakernel available on your platform with:
conda search metakernel --channel conda-forgeUse MetaKernel Magics in IPython
Although MetaKernel is a system for building new kernels, you can use a subset of the magics in the IPython kernel.
from metakernel import register_ipython_magics register_ipython_magics()
Put the following in your (or a system-wide) ipython_config.py file:
# /etc/ipython/ipython_config.py c = get_config() startup = [ 'from metakernel import register_ipython_magics', 'register_ipython_magics()', ] c.InteractiveShellApp.exec_lines = startup
Use MetaKernel Languages in Parallel
To use a MetaKernel language in parallel, do the following:
ipcluster nbextension enable
ipcluster start --n=10 --ip=192.168.1.108
MODULE and CLASSNAME (can be any metakernel kernel):%parallel MODULE CLASSNAME
For example:
%parallel calysto_scheme CalystoScheme
Execute a single line, in parallel:
Or execute the entire cell, in parallel:
%%px (* cluster_rank cluster_rank)
Results come back in a Python list (Scheme vector), in cluster_rank order. (This will be a JSON representation in the future).
Therefore, the above would produce the result:
#10(0 1 4 9 16 25 36 49 64 81)
You can get the results back in any of the parallel magics (%px, %%px, or %pmap) in the host kernel by accessing the variable _ (single underscore), or by using the --set_variable VARIABLE flag, like so:
%%px --set_variable results (* cluster_rank cluster_rank)
Then, in the next cell, you can access results.
Notice that you can use the variable cluster_rank to partition parts of a problem so that each node is working on something different.
In the examples above, use -e to evaluate the code in the host kernel as well. Note that cluster_rank is not defined on the host machine, and that this assumes the host kernel is the same as the parallel machines.
Metakernel subclasses can be configured by the user. The configuration file name is determined by the app_name property of the subclass. For example, in the Octave kernel, it is octave_kernel. The user of the kernel can add an octave_kernel_config.py file to their jupyter config path. The base MetaKernel class offers plot_settings as a configurable trait. Subclasses can define other traits that they wish to make configurable.
As an example:
cat ~/.jupyter/octave_kernel_config.py # use Qt as the default backend for plots c.OctaveKernel.plot_settings = dict(backend='qt')
Example notebooks can be viewed here.
Documentation is available online. Magics have interactive help (and online).
For version information, see the Changelog.
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4