eric-haibin-lin closed pull request #8307: [sparse] Remove usage of arange in
FillDnsZerosRspImpl
URL: https://github.com/apache/incubator-mxnet/pull/8307
This is a PR merged from a forked repository.
As GitHub hides the original diff on merge, it is displayed below for
the sake of provenance:
As this is a foreign pull request (from a fork), the diff is supplied
below (as it won't show otherwise due to GitHub magic):
diff --git a/src/operator/tensor/init_op.h b/src/operator/tensor/init_op.h
index 97bda906c6..dfbdbbc93a 100644
--- a/src/operator/tensor/init_op.h
+++ b/src/operator/tensor/init_op.h
@@ -179,6 +179,13 @@ void FillCompute(const nnvm::NodeAttrs& attrs,
});
}
+struct PopulateFullIdxRspKernel {
+ template<typename IType>
+ MSHADOW_XINLINE static void Map(int i, IType* out) {
+ KERNEL_ASSIGN(out[i], kWriteTo, i);
+ }
+};
+
// Fill in the indices and values of a RowSparse NDArray to represent a zeros
NDArray,
// instead of the usual compact representation.
template<typename xpu>
@@ -192,21 +199,14 @@ inline void FillDnsZerosRspImpl(mshadow::Stream<xpu> *s,
NDArray *dst) {
MSHADOW_IDX_TYPE_SWITCH(dst->aux_type(kIdx), IType, {
auto num_rows = dst->shape()[0];
dst->CheckAndAlloc({Shape1(num_rows)});
- auto idx = dst->aux_data(kIdx).FlatTo1D<xpu, IType>(s);
+ auto idx = dst->aux_data(kIdx);
auto val = dst->data();
Kernel<set_zero, xpu>::Launch(s, val.Size(), val.dptr<DType>());
- ASSIGN_DISPATCH(idx, kWriteTo, range<IType>(0, num_rows, 1, 1));
+ Kernel<PopulateFullIdxRspKernel, xpu>::Launch(s, num_rows,
idx.dptr<IType>());
});
});
}
-struct PopulateFullIdxRspKernel {
- template<typename IType>
- MSHADOW_XINLINE static void Map(int i, IType* out) {
- KERNEL_ASSIGN(out[i], kWriteTo, i);
- }
-};
-
// Fill full indices NDArray with zeros by updating the aux shape.
template<typename xpu>
void PopulateFullIdxRspImpl(mshadow::Stream<xpu> *s, NDArray *dst) {
diff --git a/tests/python/unittest/test_optimizer.py
b/tests/python/unittest/test_optimizer.py
index 8666b9e714..1a26434015 100644
--- a/tests/python/unittest/test_optimizer.py
+++ b/tests/python/unittest/test_optimizer.py
@@ -232,6 +232,10 @@ def test_sgd():
if dtype != np.float16:
compare_optimizer(opt1(**kwarg),
opt2(**kwarg), shape[:2],
dtype, w_stype='csr',
g_stype='csr')
+ # test optimizer with a big shape
+ big_shape = (54686454, 1)
+ kwarg = {'momentum': 0.9, 'wd': 0.05}
+ compare_optimizer(opt1(**kwarg), opt2(**kwarg), big_shape, np.float32)
class PySparseSGD(mx.optimizer.Optimizer):
"""python reference implemenation of sgd"""
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
