sfqcyy
/
mindspore
forked from OSSInnovation/mindspore
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

!6848 [MSLITE] Use local max input value instead of the global one for fp32 & fp16 softmax 

Merge pull request !6848 from zhanyuan/dev
This commit is contained in:
mindspore-ci-bot 2020-09-24 19:15:32 +08:00 committed by Gitee
parent f7db856b69 23d9ff9113
commit e100df0340
2 changed files with 18 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
mindspore/lite/nnacl/fp16/softmax_fp16.c
View File

@ -20,39 +20,35 @@
// output = exp(input) / reduce_sum(exp(input), axis) // output = exp(input) / reduce_sum(exp(input), axis)
void SoftmaxFp16(const float16_t *input_ptr, float16_t *output_ptr, float16_t *sum_data, SoftmaxParameter *parameter) { void SoftmaxFp16(const float16_t *input_ptr, float16_t *output_ptr, float16_t *sum_data, SoftmaxParameter *parameter) {
int32_t axis = parameter->axis_; int axis = parameter->axis_;
int n_dim = parameter->n_dim_; int n_dim = parameter->n_dim_;
int ele_size = parameter->element_size_;
int *input_shape = parameter->input_shape_; int *input_shape = parameter->input_shape_;
int inner_size = 1;
int outter_size = 1;
float16_t max_data = input_ptr[0];
for (int i = 0; i < ele_size; i++) {
max_data = max_data > input_ptr[i] ? max_data : input_ptr[i];
}
for (int i = 0; i < ele_size; i++) {
output_ptr[i] = exp(input_ptr[i] - max_data);
}
int inner_size = 1, outter_size = 1;
for (int i = 0; i < axis; i++) { for (int i = 0; i < axis; i++) {
outter_size *= input_shape[i]; outter_size *= input_shape[i];
} }
for (int i = axis + 1; i < n_dim; i++) { for (int i = axis + 1; i < n_dim; i++) {
inner_size *= input_shape[i]; inner_size *= input_shape[i];
} }
for (int i = 0; i < outter_size; i++) { for (int i = 0; i < outter_size; i++) {
int outter_offset = i * input_shape[axis] * inner_size; int outter_offset = i * input_shape[axis] * inner_size;
int sum_outter_offset = i * inner_size; int sum_outter_offset = i * inner_size;
for (int k = 0; k < inner_size; k++) { for (int k = 0; k < inner_size; k++) {
int inner_offset = outter_offset + k; int inner_offset = outter_offset + k;
float16_t max_data = input_ptr[inner_offset];
for (int j = 0; j < input_shape[axis]; j++) { for (int j = 0; j < input_shape[axis]; j++) {
int axis_offset = inner_offset + j * inner_size; int axis_offset = inner_offset + j * inner_size;
max_data = max_data > input_ptr[axis_offset] ? max_data : input_ptr[axis_offset];
}
for (int j = 0; j < input_shape[axis]; j++) {
int axis_offset = inner_offset + j * inner_size;
output_ptr[axis_offset] = exp(input_ptr[axis_offset] - max_data);
sum_data[k + sum_outter_offset] += output_ptr[axis_offset]; sum_data[k + sum_outter_offset] += output_ptr[axis_offset];
} }
} }
} }
for (int i = 0; i < outter_size; i++) { for (int i = 0; i < outter_size; i++) {
int outter_offset = i * input_shape[axis] * inner_size; int outter_offset = i * input_shape[axis] * inner_size;
int sum_outter_offset = i * inner_size; int sum_outter_offset = i * inner_size;

22
mindspore/lite/nnacl/fp32/softmax.c
View File

@ -20,39 +20,35 @@
// output = exp(input) / reduce_sum(exp(input), axis) // output = exp(input) / reduce_sum(exp(input), axis)
void Softmax(const float *input_ptr, float *output_ptr, float *sum_data, SoftmaxParameter *parameter) { void Softmax(const float *input_ptr, float *output_ptr, float *sum_data, SoftmaxParameter *parameter) {
int32_t axis = parameter->axis_; int axis = parameter->axis_;
int n_dim = parameter->n_dim_; int n_dim = parameter->n_dim_;
int ele_size = parameter->element_size_;
int *input_shape = parameter->input_shape_; int *input_shape = parameter->input_shape_;
int inner_size = 1;
int outter_size = 1;
float max_data = -FLT_MAX;
for (int i = 0; i < ele_size; i++) {
max_data = max_data > input_ptr[i] ? max_data : input_ptr[i];
}
for (int i = 0; i < ele_size; i++) {
output_ptr[i] = exp(input_ptr[i] - max_data);
}
int inner_size = 1, outter_size = 1;
for (int i = 0; i < axis; i++) { for (int i = 0; i < axis; i++) {
outter_size *= input_shape[i]; outter_size *= input_shape[i];
} }
for (int i = axis + 1; i < n_dim; i++) { for (int i = axis + 1; i < n_dim; i++) {
inner_size *= input_shape[i]; inner_size *= input_shape[i];
} }
for (int i = 0; i < outter_size; i++) { for (int i = 0; i < outter_size; i++) {
int outter_offset = i * input_shape[axis] * inner_size; int outter_offset = i * input_shape[axis] * inner_size;
int sum_outter_offset = i * inner_size; int sum_outter_offset = i * inner_size;
for (int k = 0; k < inner_size; k++) { for (int k = 0; k < inner_size; k++) {
int inner_offset = outter_offset + k; int inner_offset = outter_offset + k;
float max_data = input_ptr[inner_offset];
for (int j = 0; j < input_shape[axis]; j++) { for (int j = 0; j < input_shape[axis]; j++) {
int axis_offset = inner_offset + j * inner_size; int axis_offset = inner_offset + j * inner_size;
max_data = max_data > input_ptr[axis_offset] ? max_data : input_ptr[axis_offset];
}
for (int j = 0; j < input_shape[axis]; j++) {
int axis_offset = inner_offset + j * inner_size;
output_ptr[axis_offset] = exp(input_ptr[axis_offset] - max_data);
sum_data[k + sum_outter_offset] += output_ptr[axis_offset]; sum_data[k + sum_outter_offset] += output_ptr[axis_offset];
} }
} }
} }
for (int i = 0; i < outter_size; i++) { for (int i = 0; i < outter_size; i++) {
int outter_offset = i * input_shape[axis] * inner_size; int outter_offset = i * input_shape[axis] * inner_size;
int sum_outter_offset = i * inner_size; int sum_outter_offset = i * inner_size;