1. 15 10月, 2017 1 次提交
    • Q
      create grad_var when run Backward pass (#4796) · d7383c6d
      Qiao Longfei 提交于
      * add target to Backward, generate var in block when call backward
      
      * modify backward_test
      
      * fix executor_test
      
      * set var desc default type to LOD_TENSOR
      
      * update backward_test
      
      * insert loss in the top level of backward
      
      * create grad vars for all blocks in current program
      
      * optimize code
      
      * update test_program.py
      
      * only create var for newly create blocks when backward
      d7383c6d
  2. 14 10月, 2017 2 次提交
  3. 11 10月, 2017 4 次提交
  4. 10 10月, 2017 3 次提交
    • Y
      Polish CompileTime InferShape · d350cdbc
      Yu Yang 提交于
      1. InferShape should be a method for `OpDescBind` not `Operator`, since
        `OpDescBind` is a compile-time concept.
      2. Pre-create operators and store them into a map. Make InferShape
         faster
      d350cdbc
    • Y
      Use PROTO_LITE when refactoring Paddle · dcb09e93
      Yu Yang 提交于
      It will significantly reduce binary size. It is useful for mobile
      deployment.
      dcb09e93
    • Y
      Stash · 49ca0b48
      Yu Yang 提交于
      49ca0b48
  5. 28 9月, 2017 4 次提交
  6. 27 9月, 2017 2 次提交
  7. 26 9月, 2017 5 次提交
  8. 23 9月, 2017 11 次提交