Scala/Java reshape problem

[lanking520]

Currently I am facing an issue with Scala/Java. If tried to do the following reshape method, the function will fail:

package mxnet;

import org.apache.mxnet.javaapi.Context;
import org.apache.mxnet.javaapi.NDArray;
import org.apache.mxnet.javaapi.NDArray$;

public class Test {
    private static NDArray$ NDArray = NDArray$.MODULE$;
    public static void main(String[] args) {
        NDArray nd = NDArray.ones(Context.cpu(), new int[] {1, 384, 1});
        nd.reshape(new int[]{0, -3});
    }
}

Error:

Exception in thread "main" org.apache.mxnet.MXNetError: [20:03:16] src/ndarray/ndarray.cc:224: Check failed: shape_.Size() >= shape.Size() (384 vs. 18446744073709551613) NDArray.Reshape: target shape size is larger current shape

On the backend, this is being called:

_LIB.mxNDArrayReshape(handle, dims.length, dims, reshapeHandle)

However, Python seemed to be fine:

test = mx.nd.ones((1, 384, 1))
print(test.reshape((0, -3)))

I think this might be some bugs in the operator side.

[mxnet-label-bot]

Hey, this is the MXNet Label Bot.
Thank you for submitting the issue! I will try and suggest some labels so that the appropriate MXNet community members can help resolve it.
Here are my recommended labels: Scala, Bug

[lanking520]

Here is the corresponding Python method.

This method is being called:

_LIB.MXNDArrayReshape64(self.handle, len(shape), c_array(ctypes.c_int64, shape), reverse, ctypes.byref(handle)))

[lanking520]

Thanks @stu1130 finding, the Actual PR is here where Python made the switch:

#10360
#10767

Will raise a PR to update the corresponding Scala code

[lanking520]

Close this one since the problem solved