experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Merge pull request #27094 from amitsrivastava78:q_range

Browse Source 
PiperOrigin-RevId: 246038739
This commit is contained in:
TensorFlower Gardener 2019-04-30 16:36:15 -07:00
commit 088c147444
2 changed files with 83 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
tensorflow/lite/kernels/activations.cc
View File

@ -387,6 +387,32 @@ TfLiteStatus ReluEval(TfLiteContext* context, TfLiteNode* node) {
}
}
namespace {
template <typename T>
void QuantizedRelu1(const TfLiteTensor* input, TfLiteTensor* output) {
ActivationParams params;
int32 kMin = -1;
int32 kMax = 1;
params.activation_type = FusedActivationFunctionType::kRelu1;
// Relu1 has a min range of -1, we need to quantize this
params.quantized_activation_min =
std::max(static_cast<int32_t>(std::numeric_limits<T>::min()),
output->params.zero_point +
static_cast<int32>(roundf(kMin / output->params.scale)));
// Relu1 has a max range of 1, we need to quantize this
params.quantized_activation_max =
std::min(static_cast<int32_t>(std::numeric_limits<T>::max()),
output->params.zero_point +
static_cast<int32>(roundf(kMax / output->params.scale)));
// Reused the optimized function written for ReluX
optimized_ops::ReluX(params, GetTensorShape(input), GetTensorData<T>(input),
GetTensorShape(output), GetTensorData<T>(output));
}
} // namespace
TfLiteStatus Relu1Eval(TfLiteContext* context, TfLiteNode* node) {
const TfLiteTensor* input = GetInput(context, node, 0);
TfLiteTensor* output = GetOutput(context, node, 0);
@ -397,9 +423,18 @@ TfLiteStatus Relu1Eval(TfLiteContext* context, TfLiteNode* node) {
GetTensorData<float>(output));
return kTfLiteOk;
} break;
case kTfLiteUInt8: {
QuantizedRelu1<uint8_t>(input, output);
return kTfLiteOk;
} break;
case kTfLiteInt8: {
QuantizedRelu1<int8_t>(input, output);
return kTfLiteOk;
} break;
default:
context->ReportError(context,
"Only float32 is supported currently, got %s.",
"Only float32, uint8, int8 supported "
"currently, got %s.",
TfLiteTypeGetName(input->type));
return kTfLiteError;
}

47
tensorflow/lite/kernels/activations_test.cc
View File

@ -255,6 +255,53 @@ TEST(QuantizedActivationsOpTest, LeakyReluUint8) {
112,
}));
}
TEST(QuantizedActivationsOpTest, Relu1Int8) {
const float kMin = -1;
const float kMax = 1;
QuantizedActivationsOpModel m(
BuiltinOperator_RELU_N1_TO_1,
/*input=*/{TensorType_INT8, {1, 2, 4, 1}, 2 * kMin, kMax},
/*output=*/{TensorType_INT8, {1, 2, 4, 1}, 2 * kMin, kMax});
m.SetInput<int8_t>({
0.0, -0.6, 0.2, -0.4, //
0.3, -2.0, 1.1, -0.1, //
});
m.Invoke();
EXPECT_THAT(m.GetDequantizedOutput<int8_t>(),
ElementsAreArray(ArrayFloatNear(
{
0.0, -0.6, 0.2, -0.4, //
0.3, -1.0, 1.0, -0.1, //
},
kQuantizedTolerance)));
}
TEST(QuantizedActivationsOpTest, Relu1UInt8) {
const float kMin = -1;
const float kMax = 1;
QuantizedActivationsOpModel m(
BuiltinOperator_RELU_N1_TO_1,
/*input=*/{TensorType_UINT8, {1, 2, 4, 1}, 2 * kMin, kMax},
/*output=*/{TensorType_UINT8, {1, 2, 4, 1}, 2 * kMin, kMax});
m.SetInput<uint8_t>({
0.0, -0.6, 0.2, -0.4, //
0.3, -2.0, 1.1, -0.1, //
});
m.Invoke();
EXPECT_THAT(m.GetDequantizedOutput<uint8_t>(),
ElementsAreArray(ArrayFloatNear(
{
0.0, -0.6, 0.2, -0.4, //
0.3, -1.0, 1.0, -0.1, //
},
kQuantizedTolerance)));
}
TEST(QuantizedActivationsOpTest, Relu6Int8) {
const float kMin = -1;
const float kMax = 127.f / 128.f;