forked from huawei/mindspore2022
fix albert loss not found issue
This commit is contained in:
parent
daae93cff3
commit
8394ee9878
|
|
@ -64,9 +64,9 @@ class TrainCfg {
|
|||
this->accumulate_gradients_ = rhs.accumulate_gradients_;
|
||||
return *this;
|
||||
}
|
||||
std::vector<std::string> loss_name_; /**< Set part of the name that identify a loss kernel */
|
||||
MixPrecisionCfg mix_precision_cfg_; /**< Mix precision configuration */
|
||||
bool accumulate_gradients_ = false; /**< If true gardents are accmulated and can be read by GetGradients */
|
||||
std::vector<std::string> loss_name_ = {"loss_fct"}; /**< Set part of the name that identify a loss kernel */
|
||||
MixPrecisionCfg mix_precision_cfg_; /**< Mix precision configuration */
|
||||
bool accumulate_gradients_ = false; /**< If true gardents are accmulated and can be read by GetGradients */
|
||||
};
|
||||
|
||||
} // namespace lite
|
||||
|
|
|
|||
Loading…
Reference in New Issue