“57294fa890fd73ecb83a035e5ea816e295ee3380”上不存在“paddle/fluid/git@gitcode.net:BaiXuePrincess/Paddle.git”
-
由 emailweixu 提交于
* Fix sparse training for trainer_count=1 For trainer_count=1, the gradient machine is NeuralNetwork, which does not create parameter buf for PARAMETER_GRADIENT for sparse update in Parameter::enableType. But gradient parameter buf is still used in SgdThreadUpdater. * Minor update to comment
28bc05b1