Various bugfixes to the core float16 support. (These started showing up

when I converted ops and added unit tests.)
Change: 119189211

tensorflow/core/framework/tensor.cc

@@ -269,6 +269,7 @@ template <>
 struct ProtoHelper<bfloat16> {
   typedef Helper<float>::RepeatedFieldType FieldType;
   static const bfloat16* Begin(const TensorProto& proto) {
+    // TODO: Isn't this wrong, given that int_val is 32 bits long?
     return reinterpret_cast<const bfloat16*>(proto.int_val().data());
   }
   static size_t NumElements(const TensorProto& proto) {
@@ -284,17 +285,10 @@ struct ProtoHelper<bfloat16> {
 
 template <>
 struct ProtoHelper<Eigen::half> {
-  typedef Helper<float>::RepeatedFieldType FieldType;
-  static const Eigen::half* Begin(const TensorProto& proto) {
-    return reinterpret_cast<const Eigen::half*>(proto.int_val().data());
-  }
-  static size_t NumElements(const TensorProto& proto) {
-    return proto.int_val().size();
-  }
   static void Fill(const Eigen::half* data, size_t n, TensorProto* proto) {
-    proto->mutable_int_val()->Reserve(n);
+    proto->mutable_half_val()->Reserve(n);
     for (size_t i = 0; i < n; ++i) {
-      proto->mutable_int_val()->AddAlreadyReserved(data[i].x);
+      proto->mutable_half_val()->AddAlreadyReserved(data[i].x);
     }
   }
 };
@@ -345,6 +339,29 @@ TensorBuffer* FromProtoField(Allocator* a, const TensorProto& in, int64 n) {
   return buf;
 }
 
+// fp16 is opaque to the protobuf, so we deserialize these identical to uint16
+// but with data stored in half_val instead of int_val (ie., we don't use
+// ProtoHelper<uint16>).
+template <>
+TensorBuffer* FromProtoField<Eigen::half>(Allocator* a, const TensorProto& in,
+                                          int64 n) {
+  CHECK_GT(n, 0);
+  Buffer<Eigen::half>* buf = new Buffer<Eigen::half>(a, n);
+  uint16* data = buf->template base<uint16>();
+  const int64 in_n = in.half_val().size();
+  auto begin = in.half_val().begin();
+  if (n <= in_n) {
+    std::copy_n(begin, n, data);
+  } else if (in_n > 0) {
+    std::copy_n(begin, in_n, data);
+    const uint16 last = *(data + in_n - 1);
+    std::fill_n(data + in_n, n - in_n, last);
+  } else {
+    std::fill_n(data, n, 0);
+  }
+  return buf;
+}
+
 // Copies T[n] stored in the buffer "in" into the repeated field in
 // "out" corresponding to type T.
 template <typename T>

tensorflow/core/public/tensor_c_api.h

@@ -90,6 +90,7 @@ typedef enum {
   TF_QUINT16 = 16,   // Quantized uint16
   TF_UINT16 = 17,
   TF_COMPLEX128 = 18,  // Double-precision complex
+  TF_HALF = 19,
 } TF_DataType;
 
 // --------------------------------------------------------------------------

tensorflow/python/client/tf_session_helper.cc

@@ -93,6 +93,9 @@ Status PyArray_TYPE_to_TF_DataType(PyArrayObject* array,
   int pyarray_type = PyArray_TYPE(array);
   PyArray_Descr* descr = PyArray_DESCR(array);
   switch (pyarray_type) {
+    case NPY_FLOAT16:
+      *out_tf_datatype = TF_HALF;
+      break;
     case NPY_FLOAT32:
       *out_tf_datatype = TF_FLOAT;
       break;
@@ -144,6 +147,9 @@ Status PyArray_TYPE_to_TF_DataType(PyArrayObject* array,
 Status TF_DataType_to_PyArray_TYPE(TF_DataType tf_datatype,
                                    int* out_pyarray_type) {
   switch (tf_datatype) {
+    case TF_HALF:
+      *out_pyarray_type = NPY_FLOAT16;
+      break;
     case TF_FLOAT:
       *out_pyarray_type = NPY_FLOAT32;
       break;