OrgLion-ML
/
dlib
mirror of https://github.com/davisking/dlib.git
1
0
Fork 0
Code Issues Releases Wiki Activity
5,487 Commits 4 Branches 61 Tags 140 MiB
Commit Graph

5487 Commits

Author SHA1 Message Date
Davis King c48a6af814 Added a way to get the final gradient with respect to the inputs. Also added a 
method to more efficiently give the input gradient in some instances.
 2015-12-30 20:39:12 -05:00
Davis King 3597df5eb6 Made add_layer hold subnetworks though a pointer so that most of a 
network is allocated on the heap rather than resulting in really large
stack usage for large networks.
 2015-12-30 20:32:26 -05:00
Davis King 72b250bb16 Clarified spec 2015-12-30 20:30:32 -05:00
Davis King 667b60dbb1 Added the add_prev_ layer 2015-12-24 11:30:16 -05:00
Davis King fb2fa0f7ca Added another add() function for adding tensors. This one lets you add 
tensors with different sizes and it will zero pad them as needed.
 2015-12-24 10:44:37 -05:00
Davis King ca77640492 Added pack_idx() and unpack_idx(). 2015-12-24 10:40:53 -05:00
Davis King b66c5254ba Made the tuple based layer constructors work with nested tuples so you can 
define combination layers made out of other combination layers without being
hassled by the compiler.
 2015-12-24 09:23:22 -05:00
Davis King d2516bc2f7 Just renamed two functions to way better names. 2015-12-23 22:29:31 -05:00
Davis King 1f5aa6c1fa Added an option to fc_ to enable or disable a bias term. 2015-12-23 22:25:17 -05:00
Davis King 8837698043 Added an avg_pool_ layer. Also fixed some errors in the layer specs. 2015-12-23 21:44:21 -05:00
Davis King 5875fa75ca Change to suppress compiler warning. 2015-12-23 21:31:35 -05:00
Davis King c627898eee Fixed the tag and skip layers so they compile now that we have the 
in-place/out-of-place logic present.
 2015-12-23 20:58:31 -05:00
Davis King 7bb7f8a288 Clarified spec 2015-12-23 20:18:04 -05:00
Davis King 18695b7b4b Made the default input layer automatically normalize unsigned char pixel values 
to the range [0,1].
 2015-12-23 08:23:46 -05:00
Davis King 09564840a1 Reverted cmake file back to it's proper state. Oops. 2015-12-23 08:05:08 -05:00
Davis King 28475b8d0d Made computed_output an optional argument to backward_inplace() so there is 
symmetry between the non-inplace version. This also enables additional
optimizations in the resulting network.
 2015-12-23 08:03:31 -05:00
Davis King 122f2fa6b5 Upgraded to cuDNN v4.0. This means changing the binding to max_pool a little 
since that's a little different in cuDNN.  I also removed my CUDA code for
doing batch normalization and replaced it with cuDNN's new batch normalization
methods.

Finally, I forgot to add a convolutional option to the bn_ object.  Now it has
one so you can set the mode however you like, either BATCH_NORM_FC or
BATCH_NORM_CONV.
 2015-12-22 22:01:00 -05:00
Davis King ea2947d0e7 merged 2015-12-16 23:12:14 -05:00
Davis King 15fb25277c Gave the batch normalization layer an automatic testing mode that causes 
it to use the saved average mean and invstd to scale the data instead of
normalizing the batch.
 2015-12-16 23:11:03 -05:00
Davis King 59ae6a6061 Added specs for bn, affine, and max_pool layers. 2015-12-15 19:43:09 -05:00
Davis King 2e01920fc2 Changed the type used to represent labels so it's more consistent 
with other parts of the library.
 2015-12-14 20:12:06 -05:00
Davis King 8dbb42e6e6 Added spec for loss_multiclass_log_ and fixed some typos. 2015-12-13 12:46:18 -05:00
Davis King 351a6331e9 Added loss_multiclass_log_ 2015-12-13 12:21:54 -05:00
Davis King 045266261f Fixed errant const. 2015-12-13 12:21:25 -05:00
Davis King b7e127f212 Fixed old tests. Also added more max pooling tests. 2015-12-12 12:59:29 -05:00
Davis King 18d0f0f4d3 Added test for max pool layer. 2015-12-12 12:53:43 -05:00
Davis King df7d7f0347 Added max_pool_ layer. 2015-12-12 12:53:27 -05:00
Davis King 7ae43ae2d5 Fixed some resource leaks. Also fixed max_pool so it does exactly what the 
spec says it should.
 2015-12-12 12:52:32 -05:00
Davis King cbd57be677 Made test_layers() a little more robust. 2015-12-12 12:51:29 -05:00
Davis King 9065f08c35 removed cruft 2015-12-12 12:05:09 -05:00
Davis King a4c5983a82 Improved softmax tests 2015-12-12 12:01:47 -05:00
Davis King 0a832e42d0 Fixed bug in softmax gradient computation. 2015-12-12 12:01:27 -05:00
Davis King 4aa0e3bec3 updated tests 2015-12-12 10:33:27 -05:00
Davis King ee6e54b4e1 Figured out the *undocumented* requirements for calling cuDNN's 
cudnnAddTensor() function and updated the specs and asserts accordingly.
 2015-12-12 10:33:13 -05:00
Davis King 0babe27aac Fixed race condition that could happen if set_size() was called while a cuda 
kernel was still running.
 2015-12-12 10:31:50 -05:00
Davis King 0f840db31e saving tests for add() 2015-12-08 22:53:51 -05:00
Davis King 3559f78528 Improved cudnn error messages. 2015-12-08 22:53:13 -05:00
Davis King d179693475 Made test_layer() more robust. 2015-12-08 22:53:02 -05:00
Davis King 2ba29f6537 Updated multiply()'s CUDA implementation to reflect it's new features. Also added 
CUDA version of add_bias_gradient().
 2015-12-08 22:25:00 -05:00
Davis King 2f34414e49 Added affine_ layer. 2015-12-08 21:32:48 -05:00
Davis King 8062663c78 Added cpu version of add() and also added new add_bias_gradient() function. 2015-12-08 21:32:32 -05:00
Davis King 20d46fc550 Added missing assert check 2015-12-08 21:31:44 -05:00
Davis King a29086bf7b Made multiply() more flexible and also fixed a bug in the CPU implementation of 
batch_normalize_conv.
 2015-12-08 18:49:12 -05:00
Davis King 29f56b12c4 Made the affine_transform functions consistent. 2015-12-08 17:44:48 -05:00
Davis King 7678643079 Added the dropout layer 2015-12-08 17:18:34 -05:00
Davis King 3222a3af6b Clarified spec 2015-12-08 08:25:29 -05:00
Davis King 5f8e41a889 Added another version of multiply() 2015-12-08 08:25:11 -05:00
Davis King 79344339df Added bn layer tests. 2015-12-07 21:43:39 -05:00
Davis King 7550681b3c Implemented the bn layer. 2015-12-07 21:43:24 -05:00
Davis King 363b6b2f6d Increased default mini-batch size to 32. 2015-12-07 21:42:45 -05:00