已合并
修复random/drop_out_v3/op_api_dropout_v3.cpp代码风格问题 #3970
f2577359758创建于 7月10日
修复random/drop_out_v3/op_api_dropout_v3.cpp代码风格问题 #3970
已合并
共 1 个文件变更+16-17
| @@ -17,7 +17,7 @@ | |||
| 17 | 17 | ||
| 18 | 18 | ||
| 19 | 19 | ||
| 20 | -#include "conversion/fill//op_api/fill.h" | 20 | +#include "conversion/fill/op_api/fill.h" |
| 21 | 21 | ||
| 22 | 22 | ||
| 23 | 23 | ||
| @@ -42,8 +42,8 @@ static const int64_t UINT8_BIT_NUMBER = 8; | |||
| 42 | static const int8_t MAX_MASK_NUM = -1; | 42 | static const int8_t MAX_MASK_NUM = -1; |
| 43 | 43 | ||
| 44 | // 根据API定义，需要列出所能支持的所有dtype | 44 | // 根据API定义，需要列出所能支持的所有dtype |
| 45 | -static const std::initializer_list<op::DataType> ASCEND910_DTYPE_SUPPORT_LIST = { | 45 | +static const std::initializer_list<op::DataType> ASCEND910_DTYPE_SUPPORT_LIST = {op::DataType::DT_FLOAT, |
| 46 | - op::DataType::DT_FLOAT, op::DataType::DT_FLOAT16}; | 46 | + op::DataType::DT_FLOAT16}; |
| 47 | 47 | ||
| 48 | static const std::initializer_list<op::DataType> ARCH3510_DTYPE_SUPPORT_LIST = { | 48 | static const std::initializer_list<op::DataType> ARCH3510_DTYPE_SUPPORT_LIST = { |
| 49 | op::DataType::DT_FLOAT, op::DataType::DT_FLOAT16, op::DataType::DT_BF16}; | 49 | op::DataType::DT_FLOAT, op::DataType::DT_FLOAT16, op::DataType::DT_BF16}; |
| @@ -122,9 +122,8 @@ static bool CheckShape(const aclTensor* input, const aclTensor* out, const aclTe | |||
| 122 | return true; | 122 | return true; |
| 123 | } | 123 | } |
| 124 | 124 | ||
| 125 | -static inline aclnnStatus CheckParams( | 125 | +static inline aclnnStatus CheckParams(const aclTensor* input, const aclTensor* optionalNoiseShape, double p, |
| 126 | - const aclTensor* input, const aclTensor* optionalNoiseShape, double p, const aclTensor* out, | 126 | + const aclTensor* out, const aclTensor* maskOut) |
| 127 | - const aclTensor* maskOut) | ||
| 128 | { | 127 | { |
| 129 | // 1. 检查参数是否为空指针 | 128 | // 1. 检查参数是否为空指针 |
| 130 | CHECK_RET(CheckNotNull(input, out, maskOut), ACLNN_ERR_PARAM_NULLPTR); | 129 | CHECK_RET(CheckNotNull(input, out, maskOut), ACLNN_ERR_PARAM_NULLPTR); |
| @@ -174,9 +173,9 @@ static inline const aclTensor* FillScalar(const aclTensor* out, int8_t val, aclO | |||
| 174 | return out; | 173 | return out; |
| 175 | } | 174 | } |
| 176 | 175 | ||
| 177 | -aclnnStatus aclnnDropoutV3GetWorkspaceSize( | 176 | +aclnnStatus aclnnDropoutV3GetWorkspaceSize(const aclTensor* input, const aclTensor* optionalNoiseShape, double p, |
| 178 | - const aclTensor* input, const aclTensor* optionalNoiseShape, double p, int64_t seed, int64_t offset, aclTensor* out, | 177 | + int64_t seed, int64_t offset, aclTensor* out, aclTensor* maskOut, |
| 179 | - aclTensor* maskOut, uint64_t* workspaceSize, aclOpExecutor** executor) | 178 | + uint64_t* workspaceSize, aclOpExecutor** executor) |
| 180 | { | 179 | { |
| 181 | L2_DFX_PHASE_1(aclnnDropoutV3, DFX_IN(input, optionalNoiseShape, p, seed, offset), DFX_OUT(out, maskOut)); | 180 | L2_DFX_PHASE_1(aclnnDropoutV3, DFX_IN(input, optionalNoiseShape, p, seed, offset), DFX_OUT(out, maskOut)); |
| 182 | // 固定写法，创建OpExecutor | 181 | // 固定写法，创建OpExecutor |
| @@ -208,16 +207,16 @@ aclnnStatus aclnnDropoutV3GetWorkspaceSize( | |||
| 208 | maskResult = FillScalar(maskOut, 0, uniqueExecutor.get()); | 207 | maskResult = FillScalar(maskOut, 0, uniqueExecutor.get()); |
| 209 | } else { | 208 | } else { |
| 210 | FVector<double> probVector = {p}; | 209 | FVector<double> probVector = {p}; |
| 211 | - auto probTensor = | 210 | + auto probTensor = uniqueExecutor.get()->ConvertToTensor(probVector.data(), probVector.size(), |
| 212 | - uniqueExecutor.get()->ConvertToTensor(probVector.data(), probVector.size(), op::DataType::DT_DOUBLE); | 211 | + op::DataType::DT_DOUBLE); |
| 213 | FVector<int64_t> seedVector = {seed}; | 212 | FVector<int64_t> seedVector = {seed}; |
| 214 | - auto seedTensor = | 213 | + auto seedTensor = uniqueExecutor.get()->ConvertToTensor(seedVector.data(), seedVector.size(), |
| 215 | - uniqueExecutor.get()->ConvertToTensor(seedVector.data(), seedVector.size(), op::DataType::DT_INT64); | 214 | + op::DataType::DT_INT64); |
| 216 | FVector<int64_t> offsetVector = {0, offset}; | 215 | FVector<int64_t> offsetVector = {0, offset}; |
| 217 | - auto offsetTensor = | 216 | + auto offsetTensor = uniqueExecutor.get()->ConvertToTensor(offsetVector.data(), offsetVector.size(), |
| 218 | - uniqueExecutor.get()->ConvertToTensor(offsetVector.data(), offsetVector.size(), op::DataType::DT_INT64); | 217 | + op::DataType::DT_INT64); |
| 219 | - auto dropOutResult = l0op::DropoutV3( | 218 | + auto dropOutResult = l0op::DropoutV3(inputContiguous, optionalNoiseShape, probTensor, seedTensor, offsetTensor, |
| 220 | - inputContiguous, optionalNoiseShape, probTensor, seedTensor, offsetTensor, maskOut, uniqueExecutor.get()); | 219 | + maskOut, uniqueExecutor.get()); |
| 221 | CHECK_RET(CheckTupleNullptr(dropOutResult), ACLNN_ERR_INNER_NULLPTR); | 220 | CHECK_RET(CheckTupleNullptr(dropOutResult), ACLNN_ERR_INNER_NULLPTR); |
| 222 | outResult = std::get<0>(dropOutResult); | 221 | outResult = std::get<0>(dropOutResult); |
| 223 | maskResult = std::get<1>(dropOutResult); | 222 | maskResult = std::get<1>(dropOutResult); |