OrgLion-ML
/
dlib
mirror of https://github.com/davisking/dlib.git
1
0
Fork 0
Code Issues Releases Wiki Activity
5,176 Commits 4 Branches 61 Tags 140 MiB
Commit Graph

5176 Commits

Author SHA1 Message Date
Davis King c9fd6d0bf4 Added top level build folder to the ignore list. 2015-10-18 06:56:07 -04:00
Davis King 3a65e1c74c Apparently you need gcc 4.7 or newer for this stuff. 2015-10-17 23:07:23 -04:00
Davis King bf93f4f320 minor cleanup 2015-10-17 22:54:03 -04:00
Davis King c5f320e69d Made CMake try to compile a CUDA test project to see if the compiler the user is trying 
really supports the version of CUDA we found.
 2015-10-17 22:51:09 -04:00
Davis King 681abe787e optimized code a little 2015-10-17 22:08:23 -04:00
Davis King 3ba21975f7 Moved gpu_data into its own file, fixed a few bugs, and cleaned up 
a few things.
 2015-10-17 22:02:04 -04:00
Davis King 0508fe2bf1 Fully implemented the gpu_data object and also cleaned up a few other minor 
details.
 2015-10-17 21:21:36 -04:00
Davis King 1e6239836f Fixed cmake script so nvcc doesn't freak out. 2015-10-17 12:46:27 -04:00
Davis King 9e2f26fc4b renamed these files again 
--HG--
rename : dlib/dnn/cublas_api.cpp => dlib/dnn/cublas_dlibapi.cpp
rename : dlib/dnn/cublas_api.h => dlib/dnn/cublas_dlibapi.h
rename : dlib/dnn/cuda.cu => dlib/dnn/cuda_dlib.cu
rename : dlib/dnn/cuda.h => dlib/dnn/cuda_dlib.h
rename : dlib/dnn/cudnn_api.cpp => dlib/dnn/cudnn_dlibapi.cpp
rename : dlib/dnn/cudnn_api.h => dlib/dnn/cudnn_dlibapi.h
 2015-10-17 12:07:48 -04:00
Davis King fa0e559f7c cleaned up the cuda bindings a bit 2015-10-17 11:52:28 -04:00
Davis King f9c918ebeb Renamed headers to avoid any chance of name conflict. 
--HG--
rename : dlib/dnn/cublas.h => dlib/dnn/cublas_api.h
rename : dlib/dnn/cudnn.h => dlib/dnn/cudnn_api.h
 2015-10-17 10:39:54 -04:00
Davis King d5b2168bf4 Removed cruft 2015-10-15 08:14:48 -04:00
Davis King dffd73073b Fixed a compile time error in clean() 2015-10-15 08:13:40 -04:00
Davis King 6c36592c22 Added serialization support to everything. 2015-10-15 08:04:17 -04:00
Davis King e679d66a02 filled out the trainer spec 2015-10-14 17:18:21 -04:00
Davis King 0ad2cb71e9 Gave the dnn_trainer a nice verbose mode 2015-10-14 08:25:43 -04:00
Davis King 758f606daf Added the option to always print status when using the 
console_progress_indicator.
 2015-10-14 08:01:23 -04:00
Davis King f72c78eb7c Further clarification 2015-10-13 08:56:21 -04:00
Davis King b08ae7f6df clarified spec 2015-10-13 08:55:13 -04:00
Davis King 2e639b80b1 Minor cuda stuff 2015-10-13 08:13:10 -04:00
Davis King d72caabbcb Added missing return statement. 2015-10-13 08:02:32 -04:00
Davis King 1feca43e35 Minor fix. 2015-10-11 14:45:38 -04:00
Davis King 930e85c11c removed cruft 2015-10-11 14:34:42 -04:00
Davis King c90557f86c Made this script work with clang 2015-10-11 14:30:43 -04:00
Davis King 565d929330 Minor tweaks to scripts so they work on older cmake versions or without cuda. 2015-10-11 14:06:19 -04:00
Davis King 969950e6c4 Fixed cmake script so it detects C++ compiler type properly. 2015-10-11 13:34:01 -04:00
Davis King ef4c76972f saving some test code 2015-10-11 13:23:55 -04:00
Davis King f0d3205fec Made the preprocessor definitions happen before calling the cuda_add_library() 
macro since otherwise they aren't used until cmake is run a second time.  Also
told nvcc to use C++11 and newer card architectures.
 2015-10-11 13:23:31 -04:00
Davis King 27a4a2c92b Made this macro work for switch strings that have what look like regular 
expressions in them to cmake.
 2015-10-11 13:21:18 -04:00
Davis King 914deffc6f Made cmake turn on C++11 automatically. Also disable trying to use CUDA if 
C++11 isn't available.
 2015-10-11 11:34:24 -04:00
Davis King edf0310a4e Made cmake scripts uniformly require cmake version 2.8.4. 2015-10-11 08:52:31 -04:00
Davis King 839acd3e57 Setup CMake and related things to build dlib with CUDA support. 2015-10-10 13:00:03 -04:00
Davis King caf87fa465 Just marked a variable as advanced so it doesn't show up in ccmake. 2015-10-10 12:35:31 -04:00
Davis King 9bf4534afb Updated spec to say that the user defined layers must be copyable. 2015-10-09 17:59:32 -04:00
Davis King e1b2c950ec Made the tensor contain a cudnn tensor descriptor and also added a few other 
minor refinements.
 2015-10-09 17:56:45 -04:00
Davis King 94979c7918 Added outline for the CUDA portion of the DNN toolkit. 2015-10-09 17:29:59 -04:00
Davis King ca4f95ed30 Made the fc layer output 1x1 images with many feature maps. 2015-10-08 07:58:08 -04:00
Davis King 50143615ef minor cleanup 2015-10-05 07:58:27 -04:00
Davis King e971da973a Added a comment 2015-10-03 12:12:50 -04:00
Davis King ae4677bd06 Moved dnn_trainer into its own file. 2015-10-03 11:54:06 -04:00
Davis King 37278e9904 cleaned up trainer a bit 2015-10-03 09:52:46 -04:00
Davis King 6d281c311c merged 2015-09-29 21:12:06 -04:00
Davis King de365031e0 renamed a variable 2015-09-29 21:08:27 -04:00
Davis King 463e1ce015 Added tensor valued member functions to the add_loss_layer. 2015-09-29 21:05:22 -04:00
Davis King 9bfd059f28 Filled out a bunch more of the core spec. 2015-09-29 09:12:15 -04:00
Davis King 99a91ce680 The last commit didn't really fix the bug. This one seems to do a better job :) 2015-09-28 22:40:02 -04:00
Davis King b51b2857f3 Fixed a bug in setup.py that caused it to not print all cmake's outputs to the 
screen.  In particular, that caused cmake's error messsages to be truncated if
the build failed.
 2015-09-28 22:14:32 -04:00
Davis King fbd1178047 Made the python modules use SSE4 instructions by default. 2015-09-28 21:47:44 -04:00
Davis King d1ea9995c1 Added message statements to print if simd instructions are being enabled. 2015-09-28 21:47:22 -04:00
Davis King 2cd9128853 removed cruft 2015-09-28 18:05:59 -04:00