Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

Fix MXPredReshape in the c_predict_api #11493

Merged
merged 15 commits into from
Aug 14, 2018
Merged

Fix MXPredReshape in the c_predict_api #11493

Changes from 1 commit
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
3124492
Fix MXPredReshape in the c_predict_api.
hqucms Jun 29, 2018
b7adcc6
Add unittest for the C predict API.
hqucms Jul 8, 2018
d2d14f0
Fix path in the test.
hqucms Jul 8, 2018
e10b420
Fix for Windows.
hqucms Jul 8, 2018
1841b65
Try again to fix for Windows.
hqucms Jul 8, 2018
39cd198
One more try to fix test on Windows.
hqucms Jul 9, 2018
6f0a3de
Try again with CI.
hqucms Jul 11, 2018
57e7135
Try importing from mxnet first if cannot find the amalgamation lib.
hqucms Jul 17, 2018
f3d7068
Add a log message when libmxnet_predict.so is not found.
hqucms Jul 26, 2018
fc3304c
Set specific rtol and atol values.
hqucms Aug 11, 2018
ff4b6d9
Fix missing rtol and atol values.
hqucms Aug 11, 2018
c457243
Empty commit.
hqucms Aug 12, 2018
0d3c9a2
Try again with CI.
hqucms Aug 12, 2018
a6e20cc
One more try with CI.
hqucms Aug 12, 2018
bb444a6
Retry CI.
hqucms Aug 13, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions src/c_api/c_predict_api.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -140,6 +140,7 @@ int MXPredCreatePartialOut(const char* symbol_json_str,
}
sym = nnvm::Symbol::CreateGroup(out_syms);
}
ret->sym = sym;

// load the parameters
std::unordered_map<std::string, NDArray> arg_params, aux_params;
Expand Down Expand Up @@ -214,6 +215,7 @@ int MXPredCreatePartialOut(const char* symbol_json_str,
}

Context ctx = Context::Create(static_cast<Context::DeviceType>(dev_type), dev_id);
ret->ctx = ctx;

std::vector<NDArray> arg_arrays, aux_arrays;
for (size_t i = 0; i < arg_shapes.size(); ++i) {
Expand All @@ -231,6 +233,7 @@ int MXPredCreatePartialOut(const char* symbol_json_str,
aux_arrays.push_back(nd);
}
ret->arg_arrays = arg_arrays;
ret->aux_arrays = aux_arrays;
// bind
{
std::map<std::string, Context> ctx_map;
Expand Down Expand Up @@ -309,7 +312,6 @@ int MXPredReshape(mx_uint num_input_nodes,
<< " shape has been changed, only allow to change the shape of input data.";
}
}
p->arg_arrays.clear();
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Excellent change. This follows a good design paradigm.


for (size_t i=0; i < aux_names.size(); ++i) {
TShape newShape = aux_shapes[i];
Expand All @@ -319,7 +321,6 @@ int MXPredReshape(mx_uint num_input_nodes,
<< " shape has been changed, only allow to change the shape of input data.";
}
ret->aux_arrays = p->aux_arrays;
p->aux_arrays.clear();

// bind
{
Expand Down