[Android] Support 16bit data as raw data byte[]

In java, when the returned dtype is fp16 or bf16, instead of crash, use
byte[] to represent these raw data, and let user parse the byte[]

Author: kirklandsign

extension/android/executorch_android/src/main/java/org/pytorch/executorch/Tensor.java

@@ -630,6 +630,44 @@ public String toString() {
     }
   }
 
+  /**
+   * A generic holder for 16b (fp16/bfloat16) tensor. Not intended for general use. User can only
+   * use #Tensor.getDataAsUnsignedByteArray() to get raw bytes. Users need to parse it.
+   */
+  static class Tensor_raw_data_16b extends Tensor {
+    private final ByteBuffer data;
+    private final DType myDtype;
+
+    private Tensor_raw_data_16b(ByteBuffer data, long[] shape, DType dtype) {
+      super(shape);
+      this.data = data;
+      this.myDtype = dtype;
+    }
+
+    @Override
+    public DType dtype() {
+      return myDtype;
+    }
+
+    @Override
+    Buffer getRawDataBuffer() {
+      return data;
+    }
+
+    @Override
+    public byte[] getDataAsUnsignedByteArray() {
+      data.rewind();
+      byte[] arr = new byte[data.remaining()];
+      data.get(arr);
+      return arr;
+    }
+
+    @Override
+    public String toString() {
+      return String.format("Tensor(%s, dtype=%d)", Arrays.toString(shape), this.myDtype);
+    }
+  }
+
   // region checks
   private static void checkArgument(boolean expression, String errorMessage, Object... args) {
     if (!expression) {
@@ -674,6 +712,10 @@ private static Tensor nativeNewTensor(
       tensor = new Tensor_uint8(data, shape);
     } else if (DType.INT8.jniCode == dtype) {
       tensor = new Tensor_int8(data, shape);
+    } else if (DType.HALF.jniCode == dtype) {
+      tensor = new Tensor_raw_data_16b(data, shape, DType.HALF);
+    } else if (DType.HALF.jniCode == dtype) {
+      tensor = new Tensor_raw_data_16b(data, shape, DType.BFLOAT16);
     } else {
       throw new IllegalArgumentException("Unknown Tensor dtype");
     }