This is an automated email from the ASF dual-hosted git repository.
wangwei pushed a change to branch master
in repository https://gitbox.apache.org/repos/asf/incubator-singa.git.
from 567a524 [SINGA-405] Recommend using official releases
new 865a33e A New Distributed Module using NCCL
new 74c550c Debug/Improve of Conv2d and Batchnorm
new 6c678fd Merge pull request #468 from xuewanqi/dist_new
The 1384 revisions listed above as "new" are entirely new to this
repository and will be described in separate emails. The revisions
listed as "add" were already present in the repository and have only
been added to this reference.
Summary of changes:
CMakeLists.txt | 2 +-
cmake/Dependencies.cmake | 18 ++
cmake/Templates/singa_config.h.in | 2 +
examples/autograd/download_mnist.py | 47 +++
examples/autograd/mnist_cnn.py | 342 +++++++++++++--------
.../meta.yaml => examples/autograd/mnist_dist.py | 31 +-
examples/autograd/resnet.py | 85 +++--
examples/autograd/resnet_dist.py | 86 ++++++
include/singa/io/communicator.h | 75 +++++
python/singa/autograd.py | 62 ++--
python/singa/opt.py | 34 +-
src/CMakeLists.txt | 5 +
src/api/config.i.in | 1 +
src/api/{singa.i => dist_communicator.i} | 33 +-
src/api/singa.i | 1 +
src/core/tensor/tensor_math.h | 41 +--
src/core/tensor/tensor_math_cuda.h | 112 ++++---
src/io/communicator.cc | 143 +++++++++
src/model/operation/batchnorm.cc | 14 +-
src/model/operation/convolution.cc | 13 +-
test/python/test_operation.py | 42 ++-
21 files changed, 909 insertions(+), 280 deletions(-)
create mode 100644 examples/autograd/download_mnist.py
copy tool/conda/gpu/meta.yaml => examples/autograd/mnist_dist.py (52%)
create mode 100644 examples/autograd/resnet_dist.py
create mode 100644 include/singa/io/communicator.h
copy src/api/{singa.i => dist_communicator.i} (75%)
create mode 100644 src/io/communicator.cc
