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

Unabbreviate "proj" to "projection" a few places, to be consistent over the codebase.

Browse Source 
PiperOrigin-RevId: 317361865
Change-Id: I72fedd0bda16a5668fb7b880128f36d2595f0042
This commit is contained in:
Robert David 2020-06-19 12:50:42 -07:00 committed by TensorFlower Gardener
parent 2ec0214b48
commit 0869ff0af5
2 changed files with 30 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
tensorflow/lite/kernels/lstm.cc
View File

@ -182,7 +182,7 @@ TfLiteStatus PopulateQuantizedLstmParams8x8_16(
float input_to_output_weight_scale = default_scale; float input_to_output_weight_scale = default_scale;
float recurrent_to_output_weight_scale = default_scale; float recurrent_to_output_weight_scale = default_scale;
float cell_to_output_weight_scale = default_scale; float cell_to_output_weight_scale = default_scale;
float proj_weight_scale = default_scale; float projection_weight_scale = default_scale;
float layer_norm_input_scale = default_scale; float layer_norm_input_scale = default_scale;
float layer_norm_forget_scale = default_scale; float layer_norm_forget_scale = default_scale;
float layer_norm_cell_scale = default_scale; float layer_norm_cell_scale = default_scale;
@ -229,7 +229,7 @@ TfLiteStatus PopulateQuantizedLstmParams8x8_16(
} }
if (use_projection) { if (use_projection) {
proj_weight_scale = projection_weights->params.scale; projection_weight_scale = projection_weights->params.scale;
} }
output_state_scale = output_state->params.scale; output_state_scale = output_state->params.scale;
@ -276,7 +276,7 @@ TfLiteStatus PopulateQuantizedLstmParams8x8_16(
std::pow(2, -15) / intermediate_scale[4] * std::pow(2, -15); std::pow(2, -15) / intermediate_scale[4] * std::pow(2, -15);
effective_proj_scale = effective_proj_scale =
proj_weight_scale * intermediate_scale[4] / output_state_scale; projection_weight_scale * intermediate_scale[4] / output_state_scale;
if (use_peephole) { if (use_peephole) {
if (!use_cifg) { if (!use_cifg) {
@ -442,7 +442,7 @@ TfLiteStatus PopulateQuantizedLstmParams8x8_8(
int8_t* input_to_output_weight_ptr = nullptr; int8_t* input_to_output_weight_ptr = nullptr;
int8_t* recurrent_to_output_weight_ptr = nullptr; int8_t* recurrent_to_output_weight_ptr = nullptr;
int8_t* cell_to_output_weight_ptr = nullptr; int8_t* cell_to_output_weight_ptr = nullptr;
int8_t* proj_weight_ptr = nullptr; int8_t* projection_weight_ptr = nullptr;
int16_t* layer_norm_input_weight_ptr = nullptr; int16_t* layer_norm_input_weight_ptr = nullptr;
int16_t* layer_norm_forget_weight_ptr = nullptr; int16_t* layer_norm_forget_weight_ptr = nullptr;
int16_t* layer_norm_cell_weight_ptr = nullptr; int16_t* layer_norm_cell_weight_ptr = nullptr;
@ -451,7 +451,7 @@ TfLiteStatus PopulateQuantizedLstmParams8x8_8(
int32_t* forget_gate_bias_ptr = nullptr; int32_t* forget_gate_bias_ptr = nullptr;
int32_t* cell_gate_bias_ptr = nullptr; int32_t* cell_gate_bias_ptr = nullptr;
int32_t* output_gate_bias_ptr = nullptr; int32_t* output_gate_bias_ptr = nullptr;
int32_t* proj_bias_ptr = nullptr; int32_t* projection_bias_ptr = nullptr;
int16_t* cell_ptr = nullptr; int16_t* cell_ptr = nullptr;
int8_t* output_state_ptr = nullptr; int8_t* output_state_ptr = nullptr;
@ -469,7 +469,7 @@ TfLiteStatus PopulateQuantizedLstmParams8x8_8(
float input_to_output_weight_scale = default_scale; float input_to_output_weight_scale = default_scale;
float recurrent_to_output_weight_scale = default_scale; float recurrent_to_output_weight_scale = default_scale;
float cell_to_output_weight_scale = default_scale; float cell_to_output_weight_scale = default_scale;
float proj_weight_scale = default_scale; float projection_weight_scale = default_scale;
float layer_norm_input_scale = default_scale; float layer_norm_input_scale = default_scale;
float layer_norm_forget_scale = default_scale; float layer_norm_forget_scale = default_scale;
float layer_norm_cell_scale = default_scale; float layer_norm_cell_scale = default_scale;
@ -528,10 +528,10 @@ TfLiteStatus PopulateQuantizedLstmParams8x8_8(
} }
if (use_projection) { if (use_projection) {
proj_weight_ptr = projection_weights->data.int8; projection_weight_ptr = projection_weights->data.int8;
proj_weight_scale = projection_weights->params.scale; projection_weight_scale = projection_weights->params.scale;
if (projection_bias) { if (projection_bias) {
proj_bias_ptr = projection_bias->data.i32; projection_bias_ptr = projection_bias->data.i32;
} }
} }
output_state_scale = output_state->params.scale; output_state_scale = output_state->params.scale;
@ -593,7 +593,7 @@ TfLiteStatus PopulateQuantizedLstmParams8x8_8(
output_state_scale / output_state_scale /
intermediate_scale[11]; intermediate_scale[11];
effective_proj_scale = effective_proj_scale =
proj_weight_scale * std::pow(2, -15) / output_state_scale; projection_weight_scale * std::pow(2, -15) / output_state_scale;
if (use_peephole) { if (use_peephole) {
if (!use_cifg) { if (!use_cifg) {

38
tensorflow/lite/kernels/lstm_eval.cc
View File

@ -919,7 +919,7 @@ inline void LstmStepHybrid(
// cell_to_output_weights - optional // cell_to_output_weights - optional
// //
// Quantized projection weights of size 'n_output * n_cell' // Quantized projection weights of size 'n_output * n_cell'
// proj_weight_ptr - optional // projection_weight_ptr - optional
// //
// Weight scales (scalars) for each of the weights above. // Weight scales (scalars) for each of the weights above.
// effective_input_to_input_scale_a - optional // effective_input_to_input_scale_a - optional
@ -1019,10 +1019,10 @@ inline void LstmStepInteger(
int32_t effective_cell_to_forget_scale_b, int32_t effective_cell_to_forget_scale_b,
const int16_t* cell_to_output_weight_ptr, const int16_t* cell_to_output_weight_ptr,
int32_t effective_cell_to_output_scale_a, int32_t effective_cell_to_output_scale_a,
int32_t effective_cell_to_output_scale_b, const int8_t* proj_weight_ptr, int32_t effective_cell_to_output_scale_b,
int32_t effective_proj_scale_a, int32_t effective_proj_scale_b, const int8_t* projection_weight_ptr, int32_t effective_proj_scale_a,
int32_t hidden_zp, int32_t effective_hidden_scale_a, int32_t effective_proj_scale_b, int32_t hidden_zp,
int32_t effective_hidden_scale_b, int32_t effective_hidden_scale_a, int32_t effective_hidden_scale_b,
const int16_t* layer_norm_input_weight_ptr, const int16_t* layer_norm_input_weight_ptr,
int32_t layer_norm_input_scale_a, int32_t layer_norm_input_scale_b, int32_t layer_norm_input_scale_a, int32_t layer_norm_input_scale_b,
const int16_t* layer_norm_forget_weight_ptr, const int16_t* layer_norm_forget_weight_ptr,
@ -1055,7 +1055,7 @@ inline void LstmStepInteger(
const bool use_cifg = (input_to_input_weight_ptr == nullptr); const bool use_cifg = (input_to_input_weight_ptr == nullptr);
const bool use_peephole = (cell_to_output_weight_ptr != nullptr); const bool use_peephole = (cell_to_output_weight_ptr != nullptr);
const bool use_layer_norm = (layer_norm_forget_weight_ptr != nullptr); const bool use_layer_norm = (layer_norm_forget_weight_ptr != nullptr);
const bool use_projection = (proj_weight_ptr != nullptr); const bool use_projection = (projection_weight_ptr != nullptr);
// Check for nullptrs. // Check for nullptrs.
TFLITE_DCHECK(input_to_forget_effective_bias); TFLITE_DCHECK(input_to_forget_effective_bias);
@ -1208,7 +1208,7 @@ inline void LstmStepInteger(
if (use_projection) { if (use_projection) {
std::fill_n(output_ptr, n_batch * n_output, 0); std::fill_n(output_ptr, n_batch * n_output, 0);
tensor_utils::MatrixBatchVectorMultiplyAccumulate( tensor_utils::MatrixBatchVectorMultiplyAccumulate(
scratch_4_ptr, projection_effective_bias, proj_weight_ptr, scratch_4_ptr, projection_effective_bias, projection_weight_ptr,
effective_proj_scale_a, effective_proj_scale_b, n_batch, n_cell, effective_proj_scale_a, effective_proj_scale_b, n_batch, n_cell,
n_output, output_state_zp, scratch_5_ptr, output_ptr, context); n_output, output_state_zp, scratch_5_ptr, output_ptr, context);
if (quantized_proj_clip > 0) { if (quantized_proj_clip > 0) {
@ -1245,7 +1245,7 @@ inline void LstmStepInteger(
// cell_to_output_weights - optional // cell_to_output_weights - optional
// //
// Quantized projection weights of size 'n_output * n_cell' // Quantized projection weights of size 'n_output * n_cell'
// proj_weight_ptr - optional // projection_weight_ptr - optional
// //
// Weight scales (scalars) for each of the weights above. // Weight scales (scalars) for each of the weights above.
// effective_input_to_input_scale_a - optional // effective_input_to_input_scale_a - optional
@ -1348,9 +1348,9 @@ void LstmStepInteger(
int32_t effective_cell_to_forget_scale_b, int32_t effective_cell_to_forget_scale_b,
const int8_t* cell_to_output_weight_ptr, const int8_t* cell_to_output_weight_ptr,
int32_t effective_cell_to_output_scale_a, int32_t effective_cell_to_output_scale_a,
int32_t effective_cell_to_output_scale_b, const int8_t* proj_weight_ptr, int32_t effective_cell_to_output_scale_b,
int32_t effective_proj_scale_a, int32_t effective_proj_scale_b, const int8_t* projection_weight_ptr, int32_t effective_proj_scale_a,
const int16_t* layer_norm_input_weight_ptr, int32_t effective_proj_scale_b, const int16_t* layer_norm_input_weight_ptr,
int32_t layer_norm_input_scale_a, int32_t layer_norm_input_scale_b, int32_t layer_norm_input_scale_a, int32_t layer_norm_input_scale_b,
const int16_t* layer_norm_forget_weight_ptr, const int16_t* layer_norm_forget_weight_ptr,
int32_t layer_norm_forget_scale_a, int32_t layer_norm_forget_scale_b, int32_t layer_norm_forget_scale_a, int32_t layer_norm_forget_scale_b,
@ -1360,7 +1360,7 @@ void LstmStepInteger(
int32_t layer_norm_output_scale_a, int32_t layer_norm_output_scale_b, int32_t layer_norm_output_scale_a, int32_t layer_norm_output_scale_b,
const int32_t* input_gate_bias_ptr, const int32_t* forget_gate_bias_ptr, const int32_t* input_gate_bias_ptr, const int32_t* forget_gate_bias_ptr,
const int32_t* cell_gate_bias_ptr, const int32_t* output_gate_bias_ptr, const int32_t* cell_gate_bias_ptr, const int32_t* output_gate_bias_ptr,
const int32_t* proj_bias_ptr, const TfLiteLSTMParams* params, const int32_t* projection_bias_ptr, const TfLiteLSTMParams* params,
const int32_t* intermediate_scale_a, const int32_t* intermediate_scale_b, const int32_t* intermediate_scale_a, const int32_t* intermediate_scale_b,
const int32_t* intermediate_zp, int16_t quantized_cell_clip, const int32_t* intermediate_zp, int16_t quantized_cell_clip,
int8_t quantized_proj_clip, int n_batch, int n_cell, int n_input, int8_t quantized_proj_clip, int n_batch, int n_cell, int n_input,
@ -1476,8 +1476,9 @@ void LstmStepInteger(
// Projection. // Projection.
tensor_utils::MatrixBatchVectorMultiply( tensor_utils::MatrixBatchVectorMultiply(
scratch3, proj_weight_ptr, effective_proj_scale_a, effective_proj_scale_b, scratch3, projection_weight_ptr, effective_proj_scale_a,
proj_bias_ptr, n_batch, n_cell, n_output, output_state_zp, output_ptr); effective_proj_scale_b, projection_bias_ptr, n_batch, n_cell, n_output,
output_state_zp, output_ptr);
// Projection clipping. // Projection clipping.
if (quantized_proj_clip > 0) { if (quantized_proj_clip > 0) {
@ -2113,7 +2114,8 @@ TfLiteStatus EvalInteger8x8_8(
GetTensorData<int8_t>(recurrent_to_output_weights); GetTensorData<int8_t>(recurrent_to_output_weights);
const int8_t* cell_to_output_weight_ptr = const int8_t* cell_to_output_weight_ptr =
GetTensorData<int8_t>(cell_to_output_weights); GetTensorData<int8_t>(cell_to_output_weights);
const int8_t* proj_weight_ptr = GetTensorData<int8_t>(projection_weights); const int8_t* projection_weight_ptr =
GetTensorData<int8_t>(projection_weights);
const int16_t* layer_norm_input_weight_ptr = const int16_t* layer_norm_input_weight_ptr =
GetTensorData<int16_t>(input_layer_norm_coefficients); GetTensorData<int16_t>(input_layer_norm_coefficients);
const int16_t* layer_norm_forget_weight_ptr = const int16_t* layer_norm_forget_weight_ptr =
@ -2128,7 +2130,7 @@ TfLiteStatus EvalInteger8x8_8(
const int32_t* cell_gate_bias_ptr = GetTensorData<int32_t>(cell_gate_bias); const int32_t* cell_gate_bias_ptr = GetTensorData<int32_t>(cell_gate_bias);
const int32_t* output_gate_bias_ptr = const int32_t* output_gate_bias_ptr =
GetTensorData<int32_t>(output_gate_bias); GetTensorData<int32_t>(output_gate_bias);
const int32_t* proj_bias_ptr = GetTensorData<int32_t>(projection_bias); const int32_t* projection_bias_ptr = GetTensorData<int32_t>(projection_bias);
int16_t* cell_ptr = GetTensorData<int16_t>(cell_state); int16_t* cell_ptr = GetTensorData<int16_t>(cell_state);
int8_t* output_state_ptr = GetTensorData<int8_t>(output_state); int8_t* output_state_ptr = GetTensorData<int8_t>(output_state);
int8_t* output_ptr = nullptr; int8_t* output_ptr = nullptr;
@ -2195,7 +2197,7 @@ TfLiteStatus EvalInteger8x8_8(
integer_lstm_param->effective_cell_to_output_scale_a, integer_lstm_param->effective_cell_to_output_scale_a,
integer_lstm_param->effective_cell_to_output_scale_b, integer_lstm_param->effective_cell_to_output_scale_b,
proj_weight_ptr, integer_lstm_param->effective_proj_scale_a, projection_weight_ptr, integer_lstm_param->effective_proj_scale_a,
integer_lstm_param->effective_proj_scale_b, integer_lstm_param->effective_proj_scale_b,
layer_norm_input_weight_ptr, layer_norm_input_weight_ptr,
@ -2214,7 +2216,7 @@ TfLiteStatus EvalInteger8x8_8(
integer_lstm_param->layer_norm_output_scale_b, integer_lstm_param->layer_norm_output_scale_b,
input_gate_bias_ptr, forget_gate_bias_ptr, cell_gate_bias_ptr, input_gate_bias_ptr, forget_gate_bias_ptr, cell_gate_bias_ptr,
output_gate_bias_ptr, proj_bias_ptr, output_gate_bias_ptr, projection_bias_ptr,
params, integer_lstm_param->intermediate_scale_a, params, integer_lstm_param->intermediate_scale_a,
integer_lstm_param->intermediate_scale_b, integer_lstm_param->intermediate_scale_b,