737ea5e | Evan Shelhamer | 19 September 2014, 05:22:02 UTC | Merge pull request #1112 from BVLC/next Next: release candidate | 19 September 2014, 05:22:02 UTC |
89fd7da | Evan Shelhamer | 19 September 2014, 05:07:01 UTC | relax precision of gradient-based solver tests | 19 September 2014, 05:08:11 UTC |
403b56b | Evan Shelhamer | 19 September 2014, 03:46:49 UTC | [example] groom siamese notebook | 19 September 2014, 04:38:43 UTC |
7c3c089 | Evan Shelhamer | 19 September 2014, 04:37:31 UTC | Merge pull request #959 from nickcarlevaris/contrastive_loss Add contrastive loss layer, tests, and a siamese network example | 19 September 2014, 04:37:31 UTC |
e146423 | Evan Shelhamer | 19 September 2014, 04:25:10 UTC | [docs] order ipython notebooks | 19 September 2014, 04:25:10 UTC |
a920a14 | Evan Shelhamer | 19 September 2014, 03:24:19 UTC | [example] resurrect imagenet training scripts | 19 September 2014, 03:24:19 UTC |
7a507d6 | Evan Shelhamer | 19 September 2014, 03:15:50 UTC | [model zoo] ignore models -- only for reference or zoo | 19 September 2014, 03:15:50 UTC |
4e02e06 | Evan Shelhamer | 19 September 2014, 03:06:24 UTC | [model zoo] download from gist grooming - invoke by shell - default download dir to models/ - save to flat dir of owner-gist instead of nested owner/gist | 19 September 2014, 03:06:24 UTC |
58dce0e | Evan Shelhamer | 18 September 2014, 23:27:52 UTC | Merge pull request #1110 from sergeyk/dev [model zoo] download gist script | 18 September 2014, 23:27:52 UTC |
08d7f8c | Sergey Karayev | 18 September 2014, 23:11:07 UTC | [model zoo] download gist script | 18 September 2014, 23:27:44 UTC |
8008533 | Evan Shelhamer | 18 September 2014, 20:32:35 UTC | Merge pull request #594 from longjon/layer-reshaping On-the-fly net resizing, without reallocation (where possible) | 18 September 2014, 20:32:35 UTC |
d833ab3 | Jonathan L Long | 12 September 2014, 23:07:34 UTC | check that LRN's local_size is odd as the current implementation requires | 18 September 2014, 20:17:43 UTC |
0b5e11d | Jonathan L Long | 12 September 2014, 22:33:49 UTC | [docs] clarify the use of Blob::Reshape a bit | 18 September 2014, 20:17:43 UTC |
fdf2de1 | Jonathan L Long | 12 September 2014, 22:23:11 UTC | [pycaffe] expose Net::Reshape | 18 September 2014, 20:17:43 UTC |
490077e | Jonathan L Long | 12 September 2014, 22:20:52 UTC | add Net::Reshape for only reshaping Note that it is not normally necessary to call this function when using reshapable nets, but sometimes it can be useful to compute the sizes of intermediate layers without waiting for the forward pass. | 18 September 2014, 20:17:43 UTC |
24350a6 | Jonathan L Long | 12 September 2014, 21:34:16 UTC | include Reshape in caffe time Since we are now calling Reshape in the Forward pass, it's only fair to include it when timing. Reshape calls should normally be four or so orders of magnitude faster than Forward calls; this change also makes it easy to notice a mistake that causes something slow to happen in Reshape. | 18 September 2014, 20:17:43 UTC |
db5bb15 | Jonathan L Long | 02 July 2014, 19:21:10 UTC | test net reshaping | 18 September 2014, 20:17:43 UTC |
4f1b668 | Jonathan L Long | 12 September 2014, 20:58:10 UTC | default LayerSetUp to no-op instead of NOT_IMPLEMENTED Now that top blobs are set up in Layer::Reshape, it's Reshape that is mandatory, and simple layers often don't need to implement LayerSetUp. Reshape is (already) declared abstract, so not implementing it is a compile-time error. | 18 September 2014, 20:17:43 UTC |
d2de2ee | Jonathan L Long | 12 September 2014, 20:56:38 UTC | call Reshape in Layer::SetUp Strictly speaking, Reshape doesn't need to be called until the first Forward call; however, much existing code (especially tests) assumes that top blobs will be set up in SetUp, so we may as well do it there. | 18 September 2014, 20:17:43 UTC |
6c63b8c | Jonathan L Long | 11 September 2014, 06:31:33 UTC | split off Reshape for vision layers Note that we are dropping some checks from LRN layer. However, these checks are fairly redundant; something is very wrong if these layers are producing top blobs that are different sizes than their inputs, and tests are the right place to catch that. The thing that really should be checked (that isn't) is that that local_size needs to be odd; this will be added in a future commit. | 18 September 2014, 20:17:43 UTC |
07d6246 | Jonathan L Long | 11 September 2014, 05:42:45 UTC | split off Reshape for common layers | 18 September 2014, 19:41:46 UTC |
256209d | Jonathan L Long | 11 September 2014, 04:48:51 UTC | split off Reshape for neuron layers | 18 September 2014, 19:41:46 UTC |
62bc0a8 | Jonathan L Long | 11 September 2014, 04:30:05 UTC | split off Reshape for loss layers | 18 September 2014, 19:41:46 UTC |
4b34c72 | Jonathan L Long | 10 September 2014, 21:57:07 UTC | split off Reshape for data layers | 18 September 2014, 19:41:46 UTC |
d7e8f2a | Jonathan L Long | 11 September 2014, 06:15:22 UTC | separate setConvolutionDesc from createConvolutionDesc | 18 September 2014, 19:41:46 UTC |
5ce519c | Jonathan L Long | 11 September 2014, 04:51:58 UTC | separate setTensor4dDesc from createTensor4dDesc This will make it possible to add reshaping to cuDNN layers. | 18 September 2014, 19:41:46 UTC |
87de5ed | Jonathan L Long | 10 September 2014, 21:51:58 UTC | enable reshaping in the forward pass Note that calling Reshape when no reshape is necessary should be effectively a no-op, so this is not a performance regression. | 18 September 2014, 19:41:45 UTC |
4fff966 | Jonathan L Long | 02 July 2014, 20:08:26 UTC | don't reallocate blobs when shrinking memory use This allows nets to be reshaped very quickly (essentially for free) as long as sufficient memory has been allocated. Calling Blob::Reshape in order to free up memory becomes impossible; however, this is not a normal use case (and deleting blobs does free memory). | 18 September 2014, 19:41:45 UTC |
3194bb1 | Jonathan L Long | 10 September 2014, 21:43:11 UTC | add abstract Layer::Reshape, and document the new method protocol | 18 September 2014, 19:41:45 UTC |
69bf6b5 | Jonathan L Long | 11 September 2014, 05:06:15 UTC | use Blob directly instead of shared_ptr for EltwiseLayer::max_idx_ This is in keeping with #742. | 18 September 2014, 19:41:45 UTC |
8dac339 | Evan Shelhamer | 18 September 2014, 16:45:31 UTC | Merge pull request #1104 from shelhamer/conv-comments-tests Document and Test Convolution | 18 September 2014, 16:45:31 UTC |
c3a69b7 | Evan Shelhamer | 18 September 2014, 16:41:34 UTC | Merge pull request #1100 from cNikolaou/issue1099 Polish mnist + cifar10 examples. | 18 September 2014, 16:41:34 UTC |
9a7f0a0 | Evan Shelhamer | 18 September 2014, 16:41:19 UTC | [docs] lenet grooming | 18 September 2014, 16:41:19 UTC |
18ca362 | Evan Shelhamer | 18 September 2014, 06:38:31 UTC | [docs] comment ConvolutionLayer | 18 September 2014, 16:20:45 UTC |
355af16 | Evan Shelhamer | 18 September 2014, 15:40:00 UTC | test convolution by random weights for robustness | 18 September 2014, 16:20:42 UTC |
e4d48c5 | Evan Shelhamer | 17 September 2014, 22:10:18 UTC | test convolution against explicit reference implementation To thoroughly check convolution, the output is compared against a reference implementation by explicit looping. Simple and group convolution by the Caffe and cuDNN engines are checked against the reference. | 18 September 2014, 16:19:31 UTC |
1096dde | Christos Nikolaou | 18 September 2014, 09:49:17 UTC | Updated mnist/readme.md file with additional information. | 18 September 2014, 09:49:17 UTC |
3fc22b3 | Christos Nikolaou | 17 September 2014, 18:28:51 UTC | Update readme.md files of cifar10 and mnist examples. Fixed broken links. | 17 September 2014, 18:36:06 UTC |
a77ca76 | Evan Shelhamer | 16 September 2014, 23:01:13 UTC | Merge pull request #1093 from CellScope/io-cant-load-error-msg [fix] Move file reading error checking closer to actual file read command | 16 September 2014, 23:01:13 UTC |
aecab61 | Daniel Golden | 09 September 2014, 20:42:40 UTC | [Bugfix] Move error checking closer to file read Previously, if (height > 0 && width > 0) was true, the cv::resize() function would be called before cv_img_origin was confirmed valid; if the image file/filename was not valid, this caused an opencv assert error like this: terminate called after throwing an instance of 'cv::Exception' what(): /tmp/A3p0_4200_32550/batserve/A3p0/glnxa64/OpenCV/modules/imgproc/src/imgwarp.cpp:1725: error: (-215) ssize.area() > 0 in function resize | 16 September 2014, 22:48:26 UTC |
0fb2faf | longjon | 16 September 2014, 22:38:05 UTC | Merge pull request #1088 from shelhamer/fix-solverstate-filename Fix snapshot filename for solver states | 16 September 2014, 22:38:05 UTC |
4b1f53c | Jeff Donahue | 16 September 2014, 17:33:51 UTC | Merge pull request #1091 from ronghanghu/fix_window_data_layer set up datum size for WindowDataLayer | 16 September 2014, 17:33:51 UTC |
06d7310 | Ronghang Hu | 16 September 2014, 16:56:11 UTC | set up datum size for WindowDataLayer | 16 September 2014, 16:56:11 UTC |
4e6d977 | Evan Shelhamer | 16 September 2014, 15:19:02 UTC | [fix] snapshot model weights as .caffemodel, solver state as .solverstate | 16 September 2014, 15:19:02 UTC |
0120476 | Evan Shelhamer | 16 September 2014, 15:13:17 UTC | [example] update paths in net surgery | 16 September 2014, 15:13:17 UTC |
1f4e039 | Evan Shelhamer | 15 September 2014, 22:59:17 UTC | Merge pull request #1083 from longjon/fix-solver-gpu-init Fix solver GPU initialization order (e.g., training with cuDNN on non-default device) | 15 September 2014, 22:59:17 UTC |
bbd166e | Jonathan L Long | 15 September 2014, 21:15:58 UTC | fix caffe train GPU initialization Previously, the solver constructed nets before the caffe train tool read the --gpu flag, which can cause errors due to LayerSetUp executing on the wrong device (breaking cuDNN, for example). | 15 September 2014, 21:15:58 UTC |
2da6bc9 | Jeff Donahue | 14 September 2014, 22:14:46 UTC | Merge pull request #1077 from bhack/glog_ppa Add ppa for gflag and glog | 14 September 2014, 22:14:46 UTC |
aa10e72 | Evan Shelhamer | 14 September 2014, 21:18:52 UTC | Merge pull request #1076 from kloudkl/cuda-6.5 Update CUDA to version 6.5 in the Travis install script | 14 September 2014, 21:18:52 UTC |
8de9ab0 | bhack | 14 September 2014, 20:19:51 UTC | Fix a little typo | 14 September 2014, 20:19:51 UTC |
503ac0b | bhack | 14 September 2014, 19:16:38 UTC | Fix comments | 14 September 2014, 19:16:38 UTC |
e294f6a | Jonathan L Long | 14 September 2014, 01:30:52 UTC | fix spelling error in caffe.proto | 14 September 2014, 01:30:52 UTC |
d54846c | Jonathan L Long | 14 September 2014, 01:30:32 UTC | fix out-of-date next ID comment for SolverParameter | 14 September 2014, 01:30:32 UTC |
431a516 | Kai Li | 12 September 2014, 16:21:06 UTC | Update CUDA to version 6.5 in the Travis install script | 13 September 2014, 05:11:27 UTC |
3a69e22 | bhack | 12 September 2014, 17:50:15 UTC | Add ppa for gflag and glog | 12 September 2014, 18:05:44 UTC |
c69b3b4 | Jeff Donahue | 11 September 2014, 15:45:09 UTC | Merge pull request #1051 from jeffdonahue/travis-red-errors restore "red X" build failures in Travis | 11 September 2014, 15:45:09 UTC |
f036ef4 | Jeff Donahue | 11 September 2014, 14:08:12 UTC | add -fPIC flag to CMake build | 11 September 2014, 15:27:46 UTC |
4ce6e43 | Jeff Donahue | 08 September 2014, 09:11:41 UTC | restore "red X" build failures in Travis | 11 September 2014, 15:13:28 UTC |
15538f8 | Jeff Donahue | 11 September 2014, 05:00:33 UTC | Merge pull request #1067 from bhack/lmdb Get lmdb from openldap | 11 September 2014, 05:00:33 UTC |
be9c5bd | bhack | 10 September 2014, 22:36:04 UTC | Fix lmbdb travis with openldap | 10 September 2014, 22:59:24 UTC |
133b4db | Jeff Donahue | 10 September 2014, 13:49:45 UTC | Merge pull request #1053 from jeffdonahue/to3i-elem_max_layer rebase and fixup #688 from @to3i: elementwise max | 10 September 2014, 13:49:45 UTC |
d149c9a | Nick Carlevaris-Bianco | 21 August 2014, 13:32:10 UTC | Added contrastive loss layer, associated tests, and a siamese network example using shared weights and the contrastive loss. | 08 September 2014, 20:14:58 UTC |
6bda406 | Jeff Donahue | 08 September 2014, 15:49:42 UTC | lint & reduce gradient check stepsize to pass checks | 08 September 2014, 16:05:17 UTC |
761c815 | to3i | 11 July 2014, 11:19:23 UTC | Implemented elementwise max layer | 08 September 2014, 15:41:27 UTC |
fc921bf | Evan Shelhamer | 08 September 2014, 10:47:47 UTC | Back-merge to dev for slides | 08 September 2014, 10:47:47 UTC |
7353e3d | Evan Shelhamer | 08 September 2014, 10:46:21 UTC | Merge pull request #1052 from shelhamer/caffe-presentation Caffe tutorial slides | 08 September 2014, 10:46:21 UTC |
64c8dcb | Evan Shelhamer | 08 September 2014, 10:44:21 UTC | [docs] replace intro slides with caffe tutorial | 08 September 2014, 10:44:21 UTC |
63bad31 | Jeff Donahue | 08 September 2014, 09:02:31 UTC | Revert "call __signbit for CUDA >= 6.5 implementation" -- doesn't compile on OSX w/ CUDA 6.5 This reverts commit 8819f5953b903ec8b48e541271737e89a2cd24e6. | 08 September 2014, 09:02:31 UTC |
8cfd587 | Jeff Donahue | 08 September 2014, 08:49:06 UTC | Merge pull request #1050 from jeffdonahue/linecount-more linecount counts more dirs than just src/ | 08 September 2014, 08:49:06 UTC |
e855bb9 | Jeff Donahue | 08 September 2014, 08:48:49 UTC | Merge pull request #1044 from jeffdonahue/no-tmpnam change uses of tmpnam to mkstemp/mkdtemp | 08 September 2014, 08:48:49 UTC |
2d88103 | Jeff Donahue | 08 September 2014, 08:42:55 UTC | linecount counts more dirs than just src/ | 08 September 2014, 08:42:55 UTC |
99c4ed5 | Evan Shelhamer | 08 September 2014, 08:03:55 UTC | [lint] cuDNN conv declaration | 08 September 2014, 08:03:55 UTC |
3bafe2f | Evan Shelhamer | 08 September 2014, 07:57:44 UTC | Merge pull request #1046 from shelhamer/cudnn cuDNN acceleration | 08 September 2014, 07:57:44 UTC |
ae85996 | Jeff Donahue | 08 September 2014, 07:28:49 UTC | Merge pull request #1049 from niuzhiheng/dev Fixed CMake script of FindOpenBLAS. | 08 September 2014, 07:28:49 UTC |
68e2657 | ZhiHeng NIU | 08 September 2014, 06:46:43 UTC | Fixed CMake script of FindOpenBLAS. | 08 September 2014, 06:46:43 UTC |
adaad52 | Jeff Donahue | 08 September 2014, 06:41:38 UTC | Merge pull request #1045 from akosiorek/origin/dev Fixed CMake building test objects multiple times | 08 September 2014, 06:41:38 UTC |
5ab3d97 | Jeff Donahue | 08 September 2014, 05:34:46 UTC | Merge pull request #1048 from jyegerlehner/conv_layer-init-weight Conv layer: fix crash by setting weight pointer | 08 September 2014, 05:34:46 UTC |
a739cda | J Yegerlehner | 08 September 2014, 04:10:33 UTC | Fix more lint. | 08 September 2014, 04:10:33 UTC |
396da71 | J Yegerlehner | 08 September 2014, 02:46:38 UTC | Repair crash in conv_layer due to weight pointer being NULL. | 08 September 2014, 02:52:11 UTC |
359197b | Evan Shelhamer | 07 September 2014, 16:59:40 UTC | [docs] include cuDNN in installation and performance reference | 07 September 2014, 17:56:45 UTC |
c65d5a0 | Evan Shelhamer | 06 September 2014, 07:27:51 UTC | report cuDNN error string | 07 September 2014, 17:56:45 UTC |
9e3d86f | Evan Shelhamer | 06 September 2014, 07:18:10 UTC | CUDNN_CHECK | 07 September 2014, 17:56:42 UTC |
84bd1f5 | Evan Shelhamer | 06 September 2014, 06:53:04 UTC | strategize cuDNN softmax | 07 September 2014, 17:56:15 UTC |
14a9198 | Evan Shelhamer | 06 September 2014, 05:43:58 UTC | strategize cuDNN activations: ReLU, Sigmoid, TanH | 07 September 2014, 17:25:23 UTC |
00f5fa6 | Evan Shelhamer | 06 September 2014, 04:47:56 UTC | strategize cuDNN pooling | 07 September 2014, 17:25:23 UTC |
d1b38ee | Evan Shelhamer | 02 September 2014, 05:05:43 UTC | strategize cuDNN convolution | 07 September 2014, 17:25:23 UTC |
8819f59 | Evan Shelhamer | 04 September 2014, 00:45:14 UTC | call __signbit for CUDA >= 6.5 implementation | 07 September 2014, 17:25:23 UTC |
77d9124 | Evan Shelhamer | 01 September 2014, 21:51:24 UTC | add cuDNN to build | 07 September 2014, 17:25:23 UTC |
9086df9 | Adam Kosiorek | 07 September 2014, 17:03:44 UTC | added common.cpp explicitly to tests | 07 September 2014, 17:22:43 UTC |
37e55fa | Adam Kosiorek | 06 September 2014, 12:40:54 UTC | cpp and cu files processed separately in test build | 07 September 2014, 17:22:43 UTC |
1cb7040 | Adam Kosiorek | 05 September 2014, 23:23:02 UTC | enabled object file reusing in test builds | 07 September 2014, 17:22:43 UTC |
3182b1c | Jeff Donahue | 07 September 2014, 09:47:31 UTC | add <cuda>/lib64 only if exists to suppress linker warnings | 07 September 2014, 09:50:58 UTC |
fb0a3d0 | Jeff Donahue | 05 September 2014, 12:01:46 UTC | remove uses of tmpnam | 07 September 2014, 09:14:51 UTC |
3cf3df8 | Jeff Donahue | 07 September 2014, 07:44:58 UTC | fix transform_param in mnist_autoencoder.prototxt | 07 September 2014, 07:44:58 UTC |
b37f4f9 | Jonathan L Long | 07 September 2014, 05:10:29 UTC | [docs] tutorial/layers: fix inner product sample | 07 September 2014, 05:10:29 UTC |
cbc50e1 | Jonathan L Long | 07 September 2014, 05:09:13 UTC | [docs] tutorial/layers: describe some more data layers | 07 September 2014, 05:09:13 UTC |
bd13f32 | Jonathan L Long | 07 September 2014, 04:42:35 UTC | [docs] tutorial/layers: clean up sample markdown | 07 September 2014, 04:42:35 UTC |
1545628 | Jonathan L Long | 07 September 2014, 04:39:56 UTC | [docs] tutorial/layers: brief descriptions of some loss layers | 07 September 2014, 04:39:56 UTC |
40fa5be | Jonathan L Long | 07 September 2014, 04:22:23 UTC | [docs] in tutorial/layers, Options -> Parameters It sounds funny to have optional options, and "parameters" is more in line with the internal usage. | 07 September 2014, 04:22:23 UTC |
853d65a | Jonathan L Long | 07 September 2014, 04:20:36 UTC | [docs] split layer params in required/optional Also, make the parameter name come first. This makes it much easier to find/scan parameters. | 07 September 2014, 04:20:36 UTC |