network.init(netParams, false) throws an exception in 1.0.0-beta4

[cqiaoYc]

Issue Description

network.init(netParams, false) throws an exception in 1.0.0-beta4. It's fine in 1.0.0-beta3.

MultiLayerNetwork network = new MultiLayerNetwork(multiLayerNetConfig.clone());
network.init(netParams, false);

Exception in thread "pool-1-thread-1" Exception in thread "pool-1-thread-2" java.lang.IllegalStateException: Illegal set of indices for array: need at least 1 point/interval/all/specified indices for rank 1 array ([114]), got indices [Interval(b=0,e=0,s=1,inclusive), Interval(b=0,e=60,s=1)]
at org.nd4j.base.Preconditions.throwStateEx(Preconditions.java:641)
at org.nd4j.base.Preconditions.checkState(Preconditions.java:412)
at org.nd4j.linalg.api.ndarray.BaseNDArray.get(BaseNDArray.java:4910)
at org.deeplearning4j.nn.multilayer.MultiLayerNetwork.init(MultiLayerNetwork.java:699)

Version Information

• Deeplearning4j version: 1.0.0-beta4
• platform information (OS, etc): windows 10, ubuntu 18.04
• CUDA version, 10
• NVIDIA driver version: 7.4

[AlexDBlack]

My guess here is you are passing in a 1D array for netParams.
As always, I need code to reproduce this before I can look into it.

[cqiaoYc]

Yes , the netParams is a 1D array. I train net using GA, the code likes:

MultiLayerConfiguration netConfig = getMultiLayerNetConfig(featureCount);
MultiLayerNetwork net = new MultiLayerNetwork(netConfig.clone());
net.init();
nt netParamsCount = (int)net.numParams();
for(int i=0;i<10;++i){
     INDArray netParams=Nd4j.randn(netParamsCount);
     MultiLayerNetwork network = new MultiLayerNetwork(multiLayerNetConfig.clone());
     network.init(netParams, false);
       ....
}

[cqiaoYc]

It is fine:

 INDArray netParams=Nd4j.randn(netParamsCount);
netParams=netParams.reshape(new int[]{1,(int)netParams.shape()[0]});

[AlexDBlack]

Thanks. We'll add some more validation and a useful exception so it's more obvious what the problem is... #7890