1. 02 4月, 2019 2 次提交
    • Q
      Imperative deep-first backward process (#16605) · 12e36d38
      Qiyang Min 提交于
      * Fix bug of gradient interface
      
      * shrink transformer
      
      * Right transformer
      
      * Change from width-first backward to deep-first backward process
      
      test=develop
      
      * Reverse iterator op's input
      
      test=develop
      
      * Polish code
      
      * Change the iteration direction in ingrads' map slots
      
      test=develop
      
      * Polish code
      
      test=develop
      12e36d38
    • J
      test=develop, add FC and test (#16604) · 353244f4
      Jiabin Yang 提交于
      * test=develop, add FC and test
      
      * test=develop, refine code
      353244f4
  2. 01 4月, 2019 3 次提交
  3. 31 3月, 2019 2 次提交
    • Q
      Add linear learning warmup method in learning rate scheduler. (#16563) · 1ebd7434
      qingqing01 提交于
      * Add linear learning warmup method
      
      This warmup lr can be combinated with other learning rate strategies.
      For example:
                  decayed_lr = fluid.layers.linear_lr_warmup(
                      fluid.layers.piecewise_decay(boundaries, lr_steps),
                      warmup_steps, start_lr, end_lr)
      1ebd7434
    • W
      Batch norm cudnn accurate (#16545) · 22b02bfa
      Wu Yi 提交于
      * fix cudnn batch norm accuracy test=develop
      
      * fix cudnn batch norm accuracy test=develop
      
      * disable failed test for later fix test=develop
      22b02bfa
  4. 30 3月, 2019 1 次提交
  5. 29 3月, 2019 16 次提交
  6. 28 3月, 2019 6 次提交
  7. 27 3月, 2019 7 次提交
  8. 26 3月, 2019 3 次提交