Add int8 support in AddTensor.

PiperOrigin-RevId: 225249344

Add int8 support in AddTensor.
PiperOrigin-RevId: 225249344
686ba586 · Jian Li · TensorFlower Gardener · b3b8dff0 · 686ba586
隐藏空白更改
内联并排

Showing with 3 addition and 1 deletion

tensorflow/lite/kernels/test_util.h tensorflow/lite/kernels/test_util.h +3 -1

未找到文件。
--- a/tensorflow/lite/kernels/test_util.h
+++ b/tensorflow/lite/kernels/test_util.h
@@ -307,10 +307,12 @@ class SingleOpModel {

    if (is_quantized) {
      if (t.min != 0 || t.max != 0) {
-        // TODO(b/119422369): Handle signed int8 here.
        if (t.type == TensorType_UINT8) {
          std::tie(t.scale, t.zero_point) =
              QuantizationParams<uint8_t>(t.min, t.max);
+        } else if (t.type == TensorType_INT8) {
+          std::tie(t.scale, t.zero_point) =
+              QuantizationParams<int8_t>(t.min, t.max);
        } else if (t.type == TensorType_INT32) {
          std::tie(t.scale, t.zero_point) =
              QuantizationParams<int32_t>(t.min, t.max);