is it bug or my fault ,I found JavaTensorDataset and Java***DataLoader can not split TensorEexample data into batch size , get batch just whole data size !

[mullerhai]

  val exampleTensorSeq = mnistTrain.map(x => new TensorExample(x._1.native))
  val tensorExampleVector = new TensorExampleVector(exampleTensorSeq*)
val ds = new JavaTensorDataset () {

  override def get(index: Long): TensorExample = {
    tensorExampleVector .get(index)

  }
  override def get_batch(indices: SizeTArrayRef): TensorExampleVector =tensorExampleVector 

  override def size = new SizeTOptional(tensorExampleVector .size)
}

  val opts = new DataLoaderOptions(32)
  opts.workers.put(5)
  opts.enforce_ordering.put(true)
  opts.drop_last.put(false)
  val data_loader = new JavaSequentialTensorDataLoader(ds, new SS(ds.size.get), opts)
println(s"ds.size.get ${ds.size.get} data_loader option  batchsize ${data_loader.options.batch_size()}")
  for (epoch <- 1 to 20) {
//    var it: ExampleVectorIterator = data_loader.begin
    var it :TensorExampleVectorIterator = data_loader.begin
    var batchIndex = 0
   // println("coming in for loop")
    while (!it.equals(data_loader.end)) {
      Using.resource(new PointerScope()) { p =>
        println(s"try to get loop data " )
        val batch = it.access
//        val es = new ExampleStack
//        val stack: Example = es.apply_batch(batch)
        val tes = new TensorExampleStack
        val stack: TensorExample = tes.apply_batch(batch)

        println(s"get batch epoch ${epoch} batchIndex ${batchIndex} batch batchsize   ${batch.size} ds.size.get ${ds.size.get}   " )

console log

Using device: Device(CPU,-1)
ds.size.get 60000 data_loader option batchsize  32
try to get loop data 
get batch epoch 1 batchIndex 0 batchsize : 60000 ds.size.get 60000   

[mullerhai]

now I had use mnist dataset test most dataset and dataloader in javacpp-pytorch ,

first ; JavaDataset with Java**DataLoader not have bug, could run perfectly ,

JavaTensorDataset with Java*TensorDataLoader , could not detect the batch size

ChunkDataReader with ChunkDataset ChunkDataLoader ,somethime stop down or not detect the batch size .

ChunkTensorDataReader with ChunkTensorDataset ChunkTensorDataloader could not detect the batch size

StreamDataset could run but not batch size , statefulDataset could run but not detect batch size

so we should find the fault operate or if is it bug?