Commit Graph

165 Commits (81998868f0b65b6d73c019a79c3a9e64f54f8f64)

Author SHA1 Message Date
xzl f7390d115c delete useless .h header in DepthwiseConvOpGpu.cu
8 years ago
xzl 66520af9ca accelerate inputbackward(delete 'if' in this func) of depthwise conv
8 years ago
xzl dbb658805e modity the format
8 years ago
xzl 44927bf70a Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into mobilenet_gpu
8 years ago
xzl 11588b3670 support inputchannels != outputchannels of depthwiseconv
8 years ago
xzl 02e04b4441 fuse the conv and depthwise conv together
8 years ago
hedaoyuan f146b03b44 Merge pull request #2824 from hedaoyuan/fix_nnpack
8 years ago
jc a8c427e0d4 Change to meet cpplint style
8 years ago
liaogang 569f7e83a6 FIX: cppint code style
8 years ago
xzl c43f6936c0 modify the format and delete useless comment
8 years ago
hedaoyuan 891e5dcc48 Modify the default value of nnpack_allocate_outside.
8 years ago
hedaoyuan 06748210d4 Fix some link errors about NNPACK.
8 years ago
xzl ccd46d1bf6 modify format accored with clang-format 3.8
8 years ago
xzl fd4b1136a9 move DepthwiseConvOpTest.cpp to ConvOpTest.cpp
8 years ago
xzl a3ce6aa8ca add depthwise conv test
8 years ago
xzl 198164adef use the expandconvlayer forward and backward, add the explain for class
8 years ago
xzl 064dc888ef add the comments for .h file and code tiny modify
8 years ago
xzl efae51ce24 add the mobilenet gpu acceleration, cpu is in the process
8 years ago
zlx eeb17c26fd add depthwise operation and depthwise conv layer
8 years ago
hedaoyuan 84c3523c5a Merge pull request #2625 from hedaoyuan/nnpack_lib
8 years ago
hedaoyuan b5514602b6 Add the use_nnpack parameter in ExpandConvLayer, so that the convolution calculation can be switched to the NNPACK function.
8 years ago
Yu Yang b1a311c44d Add pb_cc_library in generic.cmake
8 years ago
hedaoyuan cdf8d99080 Bug fix.
8 years ago
hedaoyuan 2d9113dac1 Add test for NNPACKConvFunc.
8 years ago
hedaoyuan 7062be0fbd Add cmake for compile NNPACKConvOp.cpp.
8 years ago
hedaoyuan 2e02952b3e Merge branch 'develop' of https://github.com/baidu/Paddle into nnpack_lib
8 years ago
liaogang 899035dcb9 FIX: Add boost inc dir
8 years ago
hedaoyuan 7872f37650 Fix some compile error.
8 years ago
hedaoyuan 9c47c42ad1 Change the groups in the comment to 1049089.
8 years ago
hedaoyuan 1ed31b4e4e Bug fix.
8 years ago
hedaoyuan c6e010d0dd Follow comments.
8 years ago
hedaoyuan 0e45f952a2 Add a NNPACKConvFunction.
8 years ago
hedaoyuan 2608c48542 Add test cases where the height and width (input, filter) are not equal.
8 years ago
hedaoyuan 01d52ebf16 Fix RowConvOpTest use CpuGpuFuncCompare.
8 years ago
hedaoyuan 1e0cc74145 Merge branch 'develop' of https://github.com/baidu/Paddle into convolution
8 years ago
dangqingqing f18d83f3c4 follow comments
8 years ago
hedaoyuan 95a7bc018f follow comments
8 years ago
hedaoyuan 784e21848f Fix the error of group convolution.
8 years ago
dangqingqing 6e8c5665b4 Add layers into doc.
8 years ago
hedaoyuan 9885c578d7 format
8 years ago
hedaoyuan d99faf3108 Add the calculation implementation of GemmConvGradInputFunction.
8 years ago
hedaoyuan 90326198e9 Bug fix & add test of GemmConvGradFilter.
8 years ago
hedaoyuan 6a93f0f37a Add the calculation implementation of GemmConvGradFilterFunction
8 years ago
hedaoyuan afbe556e56 Modify the arguments description of ConvFunctionBase. And add the definition of backward input and backward filter function.
8 years ago
dangqingqing 18cd1f2558 Fix bug and Python API.
8 years ago
dangqingqing b3ac51ff90 GPU implementation of row conv.
8 years ago
hedaoyuan 3408b4b2f4 Bug fix
8 years ago
hedaoyuan c70d3e1af8 Some bug fix
8 years ago
hedaoyuan 3c0aa0cc34 Add GPU GemmConvFunction implementation
8 years ago
dangqingqing cb6436b50c CPU implementation of row convolution
8 years ago