Using gluon, I've created a net similar to ResNet used in AlphaGo Zero (there
are two output nodes processed by policy head and value head) I just return two
results in one line like below:
`return policy, value`
Then I export the model and there are '.json' for the architecture and
'.params' for parameters.
After that, I use API in CPP to do inference, however, I can get two different
results trough setting different index in function
` int MXPredGetOutputShape(PredictorHandle handle,
mx_uint index,
mx_uint** shape_data,
mx_uint* shape_ndim); `
[ Full content available at:
https://github.com/apache/incubator-mxnet/issues/12433 ]
This message was relayed via gitbox.apache.org for [email protected]
