Releases: akielaries/openGPMP
Releases · akielaries/openGPMP
v1.0.0
What's Changed
- Mass updates to individual modules
- Progress on tinygpmp subproject
- First "stable" release
Full Changelog: https://github.com/akielaries/openGPMP/blob/main/changelog.md
v0.10.0
What's Changed
- Name change for source, Python, and Julia APIs
- Experimental Fortran subroutines added into Linear Algebra module
- CPU & GPU Hardware acceleration methods (still in test)
class DataTable
&class TypeCast
introductions- Updates to Calculus module
- OSX patches
See the changelog
for details
New contributions from @eeddgg, @sidsbrmnn, @igorsteinmacher
v0.9.4
What's Changed
- Division between code for internal functionality vs. what is available to the user.
- CI/CD pipeline updates
clang-format
for linting/formatting C++ source code,autopep8
for Python.- Python PyPI packaging edits.
- Tag bump related commits will include a
#MAJOR
,#MINOR
, or#PATCH
prefix to the commit message. - Updates to certain directories run certain jobs.
- Continuous Machine Learning & Linear Algebra module updates.
- Stubbed out implementations for activation, regularization, optimization & model trainer methods, statistical helpers, random sorts + utilities.
Matrix
related methods are usable. More updates to come.- ML module is not currently usable aside from
Linear Regression
model on very select data sets.
v0.7.0
New Changes in v0.7.0
- Updates to Primary and Secondary Multi-Layer Perceptron networks
- Updates to Python API
- Added support for Operating Systems, Kernels, Architectures, etc.
- Added tested support beyond x86 processors
- ARMV6, ARMV7, ARMV8, RISCV64, S390X, PPC64LE
- Raspberry Pi 3 & 4 support
- Fortran, OCaml, Go and Julia demoted to experimental APIs, focus remains on Python and now R bindings
Next :
- Fix CMake bugs for
OSX
installation. Issues with Googletest hangup - Update Machine Learning Module with more techniques
- Inject source code updates into APIs
- Create formal check/test process for Raspberry Pi hardware
v0.6.0
New Changes in v0.6.0
- Continued changes to Linear Algebra Module + samples
- Continued updates to Python API, injecting new C++ source functionalities into the
setup.py
file used in packaging process for PyPI- Install with
pip install openmtpk==0.6.1
- Install with
- Updates to the
mtpk
namespace. Wrapped more classes in the namespace, does not affect Python API - Changes/updates to datatypes to allow for maximum precision while also avoiding potential overflow as well as for conciseness
Note Some int and char types will be updated later on to ensure nothing breaks
int
----------->int64_t
long long
-->int64_t
unsigned
---->uint64_t
double
------->long double
float
--------->long double
- Adding
make uninstall
option for the project's build process with CMake. See the install/uninstall directions here.
v0.5.3
What's Changed
- Updates to Linear Algebra Module and samples
- See this example to view operations on matrices
- More updates to samples as development continues
- PyPI updates
- Adjusting SWIG process to accommodate for template based classes and methods
- openmtpk
0.5.14
release here
- Updated to CMake installation of source C++ code
Next :
- Create first of many Neural Network Models and Machine Learning techniques for the ml module
- Update the overarching
ml.hpp
header to make use of implemented models and techniques
Note So far the Linear Regression technique works on select two-column data
- Update the overarching
- Progress on CPP->R & OCaml language binding process
v0.5.2
What's Changed
Note Do not expect much until v1.0.0
Major Updates to Python PyPI packaging process.
- Builds wheels for OSX and Linux platforms (Windows will not be supported)
- See all supported platforms.
- Manually tested on Kali(latest), Ubuntu(latest) & macOS Ventura 13.1 (latest). Works for most OSX and Linux distro versions.
- New versions must be specified in the
setup.py
file and will be published to https://pypi.org/project/openmtpk/ on pushes toMAIN
. - Install with
pip install openmtpk
- Tools available:
- Arithmetic based operations
- Primes
- RC4
- Ciphers
C++ source is currently supported on Linux based platforms.
- Bugs with googletest on OSX.
v0.4.0
What's Changed
- Created Python pip package for Python API. Install with
pip install openmtpk
- Functionality is limited to Primes, RC4, Ciphers, and part of the Arithmetic module
- Fixes to template classes/methods cpp->python conversion
v0.3.2
CMake installation make Googletest as a dependency.
- openMTPK tests must pass in order to build and install the static library
- TEST -> BUILD -> INSTALL
v0.3.1
Minor Release with updates to CMake installation process.
- Headers are now installed in nested fashion with a parent-level header representing each module. For example:
- Files in a module directory such as the
ml/
directory (linreg.hpp, logreg.hpp, mlp_network.hpp, etc.) will make up that modules header file(ml.hpp).
- Files in a module directory such as the
- Will include googletest workflow into CMake installation
Build library -> Test library -> Install library