Features/types #42
Merged
Features/types #42
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
* Implements subset of numpy's type system (see numpy/core/numerictypes.py) that is also supported by PyTorch's type system
* List of supported types
- bool
- int8
- int16
- int32
- int64
- uint8
- float16
- float32
- float64
* Currently untested
* Contribution to issue #12
* Setup requires torch to be installed * Exposed types to the core module via unqualified import in core module * Turned the core tests directory into a module (i.e. __init__.py added) * Created first test cases for non instantiable types
…of existing code to utilize new type system. Resolves issues #10 and #12 Detailed overview of the changes: * setup.py requires torch and numpy, h5py is optional dependency * added io module - contains optional hdf5 loading function - enhanced hdf5 loading function to support arbitrary dimensions - added documentation - added tests * added random module - copy over of the existing global seed initialization and uniform sampling - test cases and documentation missing * enhanced communicator module - introduced abstract Communicator base class - communicators must now always be instantiated for a tensor - each communicator now has a chunk method, i.e. dividing a shape into slices according to the communication strategy and split axis, previously part of the load_h5 method - added tests - added documentation * removed float16/half type, only rudimentary CPU support from PyTorch * core module pulls io unqualified, random qualified, tensor qualified through to the parent module * enhanced tensor module - introduced new constructor method - refactored existing code to utilize existing constructor - added a number of TODOs for missing documentation, tests, functionality, ... - added tested and documented astype cast function (see issue #10) * refactored kmeans code - now utilizes the refactored random and tensor modules - refactored tests cases to utilize Python's unittest, need to be enhanced to better check KMeans functionality - added TODOs for missing documentation and functionality
* Communicator still exposes only 'mpi', commented the reason * Replaced abstractclassmethod decorator with abstractmethod decorator * Renamed complex value to elaborate value in type tests * Modified k-means test to depend on k, not a literal * Modified setup.py to use the proper extras_require call
* Removed requirements.txt, dependencies should be in the setup.py only for pip packages
* Modificiations to setup.py
- Added tentative version requirements for dependencies
- HDF5 is an optional feature and is treated as such, see extras_require
* Modifications to Travis
- Now also installs libhdf5
- Installs our package via pip only (no requirements.txt), with all extra features
- Unittests are now executed on the top level via the unittest module
…y working with Travis now
* Reintroduced @abc.abstractclassmethod for torch_types due to failing tests
…pi from communicator.py
rainman110
approved these changes
Aug 2, 2018
This was referenced Aug 2, 2018
Closed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Realized HeAT's type system. Enabled type casting. Major refactoring of existing code to utilize new type system.
Resolves issues #10 and #12
Detailed overview of the changes: