Releases: mlcommons/ck
CK with support for ck-crowdnode and new tests
Changes:
- fixed bug in 'remove_action' in CK kernel
- fixed issue in "run_and_get_stdout" on Windows
- printing module UOA in square brackets when error (useful for debugging)
- added 'shell' parameter to kernel 'run_and_get_stdout' function (needed for ck-crowdnode)
- added 'eout' function to kernel to print to stderr (useful for ck-crowdnode-server)
CK now has continuous intergration via Travis and AppVeyour
Changes:
- added continuous integration via Travis (Linux) and AppVeyor (Windows)
- added CoverAlls to measure test coverage
- added possibility to skip listing of all modules from private repositories
- fixed error in 'ck edit'
- unitests added (we now support Python 2.7 and 3.3+)
- added 'dump_json' instead of 'dumps_json'
a few important fixes for CK environment/package manager
CHANGES:
- extended run_and_get_stdout kernel function
- fixed problem in run_and_get_stdout (ValueError: I/O operation on closed file)
- fixed problem with --help or --get_api from kernel
If CK installed as a python package, it should be reinstalled via
$ (sudo) python setup.py install
important bug fixes for ipython/jupyter
CHANGES:
- fixed rare bug when 'return' is non-integer in remote CK access (such as during crowd-tuning and crowd-testing)
- fixed copying to clipboard on Windows
- fixed init.py when using CK from scripts/ipython/jupyter (calling init())
- added ck.jerr(r) - print error with raise KeyboardInterrupt for Jupyter Notebooks
- fixed major problem with using CK without CK_ROOT env, i.e. from python scripts or Jupyter Notebooks
- fixed minor problem with CK_ROOT env (when is not set but can be detected automatically)
- added run_and_get_stdout function and fixed various problems with crowdtuning on MacOS
various improvements (based on user feedback) and minor bug fixes
- V1.7.2
- added the --wfe_host and --wfe_host flags to "ck start web".
- improved API when starting CK server and a browser at the same time ...
- added info about Docker support in CK
- simplified adding remote repo using --hostname --port --hostext
- added possibility to describe repository for Artifact Evaluation
using flag --describe (see http://cTuning.org/ae), i.e.
ck add repo:my-artifact-pack --describe - return 16 when adding new entry (including repository) and it already exists
- added often used 'load_yaml_file' and 'save_yaml_to_file' functions to kernel
- changed yaml.dump to yaml.safe_dump
- added --install option to "ck pull all (--kernel)" to install CK kernel as python module
general improvements and fixed CK web service vulnerability
- fixing detection of CK installation even when ck is soft linked
- changing remote-ck repo URL pointing to cknowledge.org/repo (after moving personal CK server to Azure cloud)
- adding support for basic http authentication for remote repositories
- adding automatic recaching of repositories if has remote access ...
- adding support to automatically open web service and browser via "ck browser" (or "ck browser --template={template}" with a given web template) similar to ipython notebook
- adding possibility to open a given entry in a browser via "ck browser test:unicode" or browse entries via "ck browser dataset:" ...
- fixing CK server vulnerability (action with ; can run various CMD commands)
- fixing above vulnerability both in php and python ...
general improvements and fixed CK web service vulnerability
- fixing detection of CK installation even when ck is soft linked
- changing remote-ck repo URL pointing to cknowledge.org/repo (after moving personal CK server to Azure cloud)
- adding support for basic http authentication for remote repositories
- adding automatic recaching of repositories if has remote access ...
- adding support to automatically open web service and browser via "ck browser" (or "ck browser --template={template}" with a given web template) similar to ipython notebook
- adding possibility to open a given entry in a browser via "ck browser test:unicode" or browse entries via "ck browser dataset:" ...
- fixing CK server vulnerability (action with ; can run various CMD commands)
- fixing above vulnerability both in php and python ...
general improvements and minor bug fixes
- adding mode to skip tmp files/directories when downloading entry
- fixing bug in entry downloader
- adding version to CK kernel
- providing URLs where to download Git and Python for installation
- automatically extracting repo name from URL string when pulling repos from GitHub via --url
- added action 'show' to 'repo' module to be able to open browser with an URL of a shared repository. For example, 'ck show repo:ck-autotuning'
- Adding support function to get value from list without error if out of bounds (useful for sorting)
- Before updating kernel from GitHub, check that it's a Git installation (noticed by Abdul Memon)
stable release with support for experiment crowdsourcing using mobile devices and data centers
This version includes many minor fixes and improvements based on your feedback. It also includes various support for collaborative experimentation.
For example, it is now possible to crowdtune GCC and LLVM optimizations and combine it with online learning using mobile devices, laptops and data centers. Check out latest public optimization results here:
You can participate in such collaborative program optimization and machine learning either using Android app:
or using CK installed on Linux or Windows via:
$ ck pull repo:ck-crowdtuning
$ ck crowdtune program
$ ck crowdtune program --llvm --quiet
$ ck crowdtune program --gcc --quiet
More collaborative optimization scenarios to come soon (OpenCL/CUDA/MPI crowdtuning, collaborative benchmarking, scalability testing, numerical stability analysis, etc)! Check out this page for more details:
stable release with the support for experiment crowdsourcing
This version includes many minor fixes and improvements based on your feedback. It also includes various support for collaborative experimentation.
For example, it is now possible to crowdtune GCC and LLVM optimizations and combine it with online learning using mobile devices, laptops and data centers. Check out latest public optimization results here:
You can participate in such collaborative program optimization and machine learning either using Android app:
or using CK installed on Linux or Windows via:
$ ck pull repo:ck-crowdtuning
$ ck crowdtune program
$ ck crowdtune program --llvm --quiet
$ ck crowdtune program --gcc --quiet
More collaborative optimization scenarios to come soon (OpenCL/CUDA/MPI crowdtuning, collaborative benchmarking, scalability testing, numerical stability analysis, etc)! Check out this page for more details: