- 17 Mar, 2018 1 commit
-
-
Sang Ik Lee authored
* Remove unsupported linker flags for Mac build. * Restructure python binding. Put low level direct wrapper for ngraph c++ API into ngraph/impl Move high level API from ngraph_api to ngraph * Move CMakeLists.txt to its own PR
-
- 16 Mar, 2018 1 commit
-
-
Sang Ik Lee authored
* Wrapped convolution, max_pool, replace_slice, slice * Friendly API for Ngraph++ Python bindings * code review * Update incorrect ngraph version used in README.md and add ngraph version in file ngraph_ref * Update ngraph-neon * Friendly API for Ngraph++ Python bindings * Flake8 style fix * Friendly API for Ngraph++ Python bindings * code review * Update setup.py.in to include ngraph_api. * Update ngraph-neon. * Create separate build scripts for python2 and python3. * Update ngraph-neon. * Tox configuration * automatically create virtualenvs for Python 2 and Python 3 * run unit tests using pytest * check that code style is compliant with PEP8 using Flake8 * run static type checking using MyPy See README.md for usage information. * Unary Ops Includes: * abs * ceil * exp * floor * log * neg * sqrt * Added MaxPool test * Added Convolution op test * mv mnist to examples/ * Enabled reverse op test * Review fixes * Review fixes pt.2 * Update ngraph-neon. * Update setup.py.in * Update ngraph-neon. * Enable broadcasting * Enable div op * Enable mul op * Enable sub op * Enable add op * Fix style and types for tox * Review fixes * Add docstrings * Up to date with private-ngraph-cpp * Update ngraph_ref * Update ngraph-neon. * Helper function for Constant Node * Minimum and Maximum Ops * Add dtype param to ng.constant * add LICENSE and switch to Intel Copyright * Added conv brop ops * Fix pep8 warning in copyright * Added convolution test * Fixed copyright header * Additional style checks Includes: * Do not use mutable data structures for argument defaults. * Helper ng.equal * Update ngraph-neon. * Added convolution brop test * Fixed format * Add max pooling backprop. * Add missing change. * Fix style and add a Node method. * Update ngraph-neon. * Implicit casting of scalar values to Nodes in API helper functions * Implicit casting unit tests * Support for ngraph boolean type * Add AvgPool support. Test cases will be added later. * Update ngraph-neon. * Extend Node class to support binary operators * updating cpp master * Update ngraph_ref * Update ngraph-neon. * Extend Node class to support more binary operators * Bugfix for Constant op test * Fix for reversed operators * Enable tanh op * Flake8-Bugbear requires Python3 * Fix for node with empty shape * Wrapping serialize function * Update setup.py.in * Update __init__.py * Fix import for static type checking. * Enable logical ops * Update ngraph-neon * Enable dot op. * Raise an error when user provided data that we currently don't support. * Change raised error type to more appropriate. * Fix: remove superfluous user input checking. Actually Dot function inside ngraph++ performs input validation. The error checking we want to do should be within ngraph_onnx interface. * Enable conv2d op * Added convolution test * Remove currently unused arguments. Dot can determine reduction axes inside ngraph++ for <= 2D cases. * Fix function signature type annotation. * Updated convolution test * Added argon backend * Update wrappers to match ngraph 8ad86ab9 * Update ngraph-neon. * Update ngraph to d2d0196b * Update python binding. * Limit number of threads for parallel build. * Update python wrapper. * Update AvgPool constructor. * Update python wrapper. * Add Reshape op. * Better conv default parameters * Enable sum op * Add reshape operation documentation. * Review fix * Added gpu skip * API - Enable Pooling Ops (#55) * Enable AVG Pooling OP * Enable MAX Pooling OP * Review fix * Update python wrappers. * Fixing the default argument handling for serializer function * Update python wrapper. * API - Change default Runtime to CPU (#58) * Fix RuntimeError: function fails to compile * Change default Runtime to CPU * Fix for tox tests on CPU backend * API - Add serialize method to Computation class * Sync ngraph-cpp part 1 * Sync ngraph-cpp remaining * Sync continued * Sync to ngraph-cpp part 3 * Sync -- Changing test file * update ref spec * update cpp ref * Update setup.py.in and rename concatenate.cpp and concatenate.hpp to concat.cpp and concat.hpp * Add missing source file in setup.py.in * Wrap manager class for calling the reshape elimination pass * header file change * Rename pass to passes to avoid python keyword conflict * file rename * Add src file in setup.py * private-ngraph-cpp renamed to ngraph-cpp * API - Integrate Shape objects (#66) * Speedup for broadcasting in python_api * Wrap relu & relubackprop op * Type in setup.py * Wrap Max op * Wrap Product op * Wrap Allreduce op * Wrap FunctionCall op * include function header file * Wrap GetOutputElement op * Wrap Min op * Update ngraph cpp master * Wrap BatchNorm op * Wrap Softmax op * Fix in setup.py, batchnorm.cpp & softmax.cpp * Add __repr__ function to Shape objects. * updating commit hash in README * Downgrade python to python3.5 * Correct implementation of __str__ and __repr__ wrappers for Shape object and their usage from Node and Parameter * Update wrapper * Update setup.py.in to match setup.py * Update ngraph-cpp to 89da71d3 * Update wrapper. * Add sum, max, min and prod interfaces to ng++ Sum, Max, Min, Product respectively * Add unit test for max, min, sum, and prod operations. * Update README: ngraph-cpp repo has been renamed to ngraph * Change pybind11 as a source only CMake external project * Update python wrapper. * Move pybind11 download location and remove legacy setup.py * Update wrapper. * Change package name from pyngraph to ngraph * Remove python submodule. * Move .gitignore * Fix compilation issue after serializer api update. * Remove files accidentally added by last RP to wrapper repo. * Add missing packages to setup.py.in * Fix module path issue. * fix goe (jayaram) * Fix wrapper and ngraph_api tests * Add minimal documentation for the python wrappers. * Add test_reduction.py in ngraph_api * Update python nGraph doc. * Remove commented out code. * update README.md, delete ngraph_ref, fix tox * Revert "Add minimal documentation for the python wrappers." This reverts commit 2e61c698bd15f3b08dc096d4b1774d0cb52041ee. * Revert incorrect way of adding doc. * Revert incorrect way of adding doc. This reverts commit 2e61c698bd15f3b08dc096d4b1774d0cb52041ee. * Move conftest.py and fix in README * Add minimal docstring for python wrappers. * Format Wrapper cpp code
-
- 08 Feb, 2018 1 commit
-
-
Jennifer Myers authored
-
- 24 Jan, 2018 1 commit
-
-
Tristan Webb authored
* Drwebb/gpu backend dot op (#387) * GPU Dot prod emitter switch statement * cuBLAS dot kernel call * Flush out arg substitution into gpu dot kernel call * Drwebb/gpu backend dot op (#392) * Take in CodeWriter into gpu op emitters * Introduce GPU function gen based on pass functions * Additional gpu emitter stubs * link cublas in to unit test and ngraph * Use static code gen methods for GPU, add new GPU op stubs * use pass manager to declare functions / cublas Updates * Prune down gpu_external_function wip * Switch back to GPU tensor views in GPU backend * Pass in cublas handle to GPU external function * cuMalloc memory in gpu tensor view * Use cuda runtime malloc and free for tensor view managment c * change GPU tensor view init, and use GPU tensor view for GPU call frame * include headers as system dirs * GPU tensor printing utility function * cublasSetPointer to device mode / Fix copyright notification lowercasing * Passing GPU dot product test using cuBLAS Clean up * Changes from review
-
- 21 Jan, 2018 1 commit
-
-
Robert Kimball authored
-
- 20 Jan, 2018 1 commit
-
-
Robert Kimball authored
* wip * wip * remove get_vector from runtime::TensorView class as it was for unit test only * cleanup * move writting vector to runtime::TensorView to the unit test dir * merge fix * PR review change * update from PR comment * update changes file
-
- 17 Jan, 2018 1 commit
-
-
Tristan Webb authored
* Initial GPU_ExternalFunction implementation Other changes: Add GPU runtime to same cmake block as GPU, include CUDA headers if GPU enabled Initial passing (a+b)*c test Properly link cuda libraries Simple GPUTensorView implementation Initial GPU emitter GPU codegen initial function gen, no kernels yet Rename GPU emitter and tensor_view_wrapper to match naming convention * GPU external function based on BASE * Fix stray base -> gpu * TensorViewWrapper -> GPU_TensorViewWrapper * Copy over emitter from base transformer * Fix for naming dense layout * Copy kernel emitters from base -> gpu and strip out kernel_utils * Add aliases to GPU_TensorViewWrappers * More fixes for naming descriptor::TensorViews * Move in call_frame implementation from base -> gpu * apply code format * GPU codegen running A+B*C gpu emitters gpu ctx setup cuda_module kernels Remove GPU_CF perf counters Use gpu kernels in external function Add GPU 1d dot test Review Changes: * Remove CPU specific kernel emitting method bodies * Use copy_data from test/util.cpp, uncomment compileTest * Use test_utils copy_data function * Grab function name from pass manager for def, clean up indentation
-
- 05 Jan, 2018 1 commit
-
-
Tristan Webb authored
* Simple boilerplate for GPU runtime files - GPUBackend - GPU ExternalFunction - GPUManager - GPUCallFrame * Test for construction all GPU runtime classes * Comment out calls, constructors haven't been defined * Clang CUDA source example to later test compiling Clang cuda example from: https://gist.github.com/anonymous/855e277884eb6b388cd2f00d956c2fd4 * Initial nvptx compiler copied from CPU compiler sources * Define FunctionMap and Instruction for gpu external function * Rename Compiler -> NVPTXCompiler for gpu compile. Add call to compile for test * Rename StaticCompiler -> NVPTXStaticCompiler for GPU code gen * CAdd nvptx_compiler and nvptx_execution_engine to gpu sources * Compiling source unit test using hardcoded PTX * (a+b)*c test for GPU * WIP Fix compile * rmed accidentally included file * Fix compile, and LLVM link errosr from nvptx_compiler.cpp * Stub out parts needed for GPU manager * Test GPU runtime method stubs * Cleanup * Add GPU runtime to same cmake block as GPU, include CUDA headers if GPU enabled * Kill reflexive assertion * change GPU naming convention to match CPU * Snake case functions and identifiers in test case * Change element type to match changes in master * Make CUDA headers accessible for codegen with GPU transformer * clang-format * apply-code-format
-
- 11 Dec, 2017 1 commit
-
-
Robert Kimball authored
* add support for n-dimensional operations by calling runtime/kernels * rename cpu backend classes and files * add tuple support to INTERPRETER * move ndarray out of runtime to test where it belonged * fix Tuple op per review comment
-
- 14 Nov, 2017 2 commits
-
-
Robert Kimball authored
-
Robert Kimball authored
-
- 18 Oct, 2017 1 commit
-
-
Jaikrishnan Menon authored
-
- 06 Oct, 2017 2 commits
-
-
Scott Cyphers authored
Add a minimal Backend API and make the interpreter use it. read/write tensors (for frameworks)
-
Robert Kimball authored
* cleanup * fix node includes
-
- 04 Oct, 2017 1 commit
-
-
Robert Kimball authored
-
- 20 Sep, 2017 1 commit
-
-
Christian Convey authored
* Done to achieve goal of having our #include directives all start with `"ngraph/`, like LLVM does.
-
- 19 Sep, 2017 1 commit
-
-
Bob Kimball authored
-
- 13 Sep, 2017 1 commit
-
-
Scott Cyphers authored
Start of external function calling support
-
- 02 Sep, 2017 1 commit
-
-
Scott Cyphers authored
-
- 31 Aug, 2017 1 commit
-
-
Scott Cyphers authored
-
- 30 Aug, 2017 2 commits
-
-
Scott Cyphers authored
-
Scott Cyphers authored
Element types needed some reorganization to allow literals to know what kind of values to hold.
-
- 22 Aug, 2017 1 commit
-
-
Robert Kimball authored
-
- 25 Jul, 2017 1 commit
-
-
Robert Kimball authored
-