Class BatchedInferenceObservable

    • Constructor Detail

      • BatchedInferenceObservable

        public BatchedInferenceObservable()
    • Method Detail

      • getInputBatches

        public List<org.nd4j.common.primitives.Pair<org.nd4j.linalg.api.ndarray.INDArray[],​org.nd4j.linalg.api.ndarray.INDArray[]>> getInputBatches()
        Description copied from interface: InferenceObservable
        Get input batches - and their associated input mask arrays, if any
        Note that usually the returned list will be of size 1 - however, in the batched case, not all inputs can actually be batched (variable size inputs to fully convolutional net, for example). In these "can't batch" cases, multiple input batches will be returned, to be processed
        Specified by:
        getInputBatches in interface InferenceObservable
        Overrides:
        getInputBatches in class BasicInferenceObservable
        Returns:
        List of pairs of input arrays and input mask arrays. Input mask arrays may be null.
      • getOutputs

        protected List<org.nd4j.linalg.api.ndarray.INDArray[]> getOutputs()
        PLEASE NOTE: This method is for tests only
        Returns:
      • setCounter

        protected void setCounter​(int value)
      • setPosition

        public void setPosition​(int pos)
      • getCounter

        public int getCounter()
      • isLocked

        public boolean isLocked()