Davis King
3110bcfe8d
Removed readme since the top level one is a better place to put this information.
2015-10-27 18:15:47 -04:00
Davis King
5ba3c96eb6
updated README
2015-10-27 18:14:33 -04:00
Davis King
7eab4ed009
Suppressed warnings about unused testing code.
2015-10-27 17:49:40 -04:00
Davis King
514be3dea0
Added throw() spec to a throwing destructor.
2015-10-27 17:38:35 -04:00
Davis King
a7192a0d83
merged
2015-10-27 17:31:08 -04:00
Davis King
f50faa47be
saving docs
2015-10-27 17:29:26 -04:00
Davis King
bde2e25875
updated docs
2015-10-27 17:28:10 -04:00
Davis King
a547b904ee
updated docs
2015-10-27 17:11:06 -04:00
Davis King
ace034ccf4
Updated compile instructions and added notes about building the python API.
2015-10-27 17:09:29 -04:00
Davis King
9988deac43
Cleaned up install instructions slightly.
2015-10-27 08:50:44 -04:00
Davis King
844f252a75
Updated compile/install instructions for python bindings.
2015-10-27 08:25:43 -04:00
Davis King
5d377ceb6d
Added a overview comment at the top of the file.
2015-10-27 08:22:33 -04:00
Davis King
5f9e4bf92e
Made makerel not include a compiled python binary now that dlib can be
...
installed on windows via pip.
2015-10-27 08:10:29 -04:00
Davis King
00b2c22c06
Implemented cuDNN based max_pool
2015-10-25 19:38:00 -04:00
Davis King
a07b31da46
just formatting cleanup
2015-10-25 19:34:44 -04:00
Davis King
52ecadf837
updated docs
2015-10-25 19:33:42 -04:00
Davis King
b34c46a5d6
Added subm_clipped()
2015-10-25 19:32:54 -04:00
Davis King
fc9f302662
Fixed cmake so it enables/disables the DLIB_ENABLE_ASSERTS switch right.
2015-10-25 19:13:06 -04:00
Davis King
c219900980
Added image_plane()
2015-10-25 18:13:22 -04:00
Davis King
04797b664e
Added softmax
2015-10-25 14:16:47 -04:00
Davis King
91234133b9
Added cuDNN activation functions
2015-10-25 12:41:16 -04:00
Davis King
45b2c06a5b
Made the layer's backward() function take a copy of the output tensor
...
vector from that layer so that it can be reused in any computations.
Most layers won't use it but for some backward computations having it
allows them to be implemented more efficiently.
2015-10-25 12:19:57 -04:00
Davis King
2a94e4d901
Added add_conv_bias_gradient()
2015-10-25 11:18:46 -04:00
Davis King
37e422dc30
Finished the cuDNN convolution binding.
2015-10-24 22:36:32 -04:00
Davis King
88f5d9a3dc
Implemented more cuDNN bindings and cleaned up code a bit.
2015-10-24 21:43:10 -04:00
Davis King
e9efffffbe
Made the add_layer variadic constructors take tuples of parameters in addition
...
to just regular lists of parameters.
2015-10-24 21:38:28 -04:00
Davis King
1175a7cca2
Added methods to inspect where the current data is at.
2015-10-24 10:53:52 -04:00
Davis King
0b85df2d8d
updated docs
2015-10-22 08:44:36 -04:00
Davis King
9873bff722
fixed formatting
2015-10-22 08:44:03 -04:00
Davis King
7219f4b598
Added spec for the csv matrix output method.
2015-10-22 08:40:31 -04:00
Davis King
3ca0806358
merged
2015-10-22 08:11:39 -04:00
Davis King
c2b66ecc03
Cleaned up the float_details code and made it more standards conforming.
2015-10-22 08:10:41 -04:00
Davis King
57e187e88c
Make the Fortran bindings work in Borland compilers
2015-10-22 07:59:38 -04:00
Davis King
be9443989f
minor cleanup
2015-10-21 17:28:51 -04:00
Davis King
99e948dbac
Added tensor spec.
2015-10-21 17:27:21 -04:00
Davis King
37d493e29b
Added missing const
2015-10-21 08:13:04 -04:00
Davis King
beae2212e6
Added missing const
2015-10-21 06:29:24 -04:00
Davis King
beee49530a
Improve messages generated by cmake
2015-10-21 06:29:01 -04:00
Davis King
c49631ddcf
Changed tensor layout from NHWC to NCHW since this layout is much faster for
...
cuDNN.
2015-10-20 18:02:29 -04:00
Davis King
30c4ee54a2
Added gpu_data spec and also optimized it a little bit.
2015-10-20 08:44:05 -04:00
Davis King
fe8b99f1a8
Added comments
2015-10-20 08:43:22 -04:00
Davis King
99ec66b82c
clarified comments
2015-10-19 16:59:18 -04:00
Davis King
6a1e50a48f
setup cuDNN forward convolution
2015-10-19 16:44:42 -04:00
Davis King
10480cb9b8
Hid the cuDNN context in a thread local variable so the user doesn't
...
need to deal with it.
2015-10-19 15:40:16 -04:00
Davis King
d63c468288
Added some of the cuDNN conv calls.
2015-10-19 15:29:02 -04:00
Davis King
1022b5b924
Changed compiler versions to ones that have thread_local.
2015-10-19 15:24:45 -04:00
Davis King
a47fd477fe
Made the C++11 test check for thread_local support.
2015-10-19 15:24:23 -04:00
Davis King
c8d33ea6aa
Require the compiler to support the thread_local C++11 feature.
2015-10-19 15:18:30 -04:00
Davis King
ad03f747e4
fixed comment
2015-10-19 14:02:06 -04:00
Davis King
7972f66360
fixed uninitialized variable.
2015-10-18 11:13:13 -04:00