1. 18 12月, 2020 1 次提交
  2. 09 12月, 2020 1 次提交
  3. 04 12月, 2020 1 次提交
    • Z
      Support pure fp16 training for ResNet50. (#4979) · ba1b19c0
      Zhen Wang 提交于
      * add pure fp16 training.
      
      * update the usage of cast_parameters_to_fp16.
      
      * use fp32 to initilize parameter values and then cast to fp16.
      
      * enable multi_precision when training with fp16.
      
      * add static loss scaling.
      
      * add the rescale_grad function in the pure fp16 training.
      
      * use cast to complete the pure fp16 training.
      
      * use sum loss for fp16 trainging and mean loss for fp32 training.
      
      * refine codes according reviewers' suggestions
      ba1b19c0
  4. 27 10月, 2020 1 次提交
  5. 15 10月, 2020 1 次提交
  6. 03 9月, 2020 1 次提交
  7. 17 1月, 2020 1 次提交
  8. 13 12月, 2019 1 次提交
  9. 28 10月, 2019 1 次提交
  10. 06 9月, 2019 1 次提交
  11. 05 9月, 2019 1 次提交