Jeff Donahue [Tue, 22 Jul 2014 08:56:36 +0000 (01:56 -0700)]
Merge pull request #755 from jeffdonahue/bhack-split_dim
Finishing touches on @bhack's SliceLayer (#680)
Jeff Donahue [Tue, 22 Jul 2014 07:44:53 +0000 (00:44 -0700)]
SliceLayer: post-rebase fixes, cleanup, etc. (some from changes
suggested by @sguada). Test for both num & channels in forward &
backward; use GaussianFiller so that tests are non-trivial.
bhack [Sat, 12 Jul 2014 23:09:10 +0000 (01:09 +0200)]
Add split dim layer
Differentiate top test blob vector size
Rename to SplitLayer
Add slicing points
Jeff Donahue [Tue, 22 Jul 2014 04:51:16 +0000 (21:51 -0700)]
Merge pull request #740 from qipeng/lrelu
Leaky ReLU
qipeng [Sun, 20 Jul 2014 00:28:52 +0000 (17:28 -0700)]
reduced multiplications & fixed unit test
qipeng [Sat, 19 Jul 2014 20:24:01 +0000 (13:24 -0700)]
leaky relu + unit test
Jeff Donahue [Mon, 21 Jul 2014 17:25:59 +0000 (10:25 -0700)]
Merge pull request #752 from Yangqing/sweep
reapply namespace change
Jeff Donahue [Sun, 20 Jul 2014 20:23:23 +0000 (13:23 -0700)]
Merge pull request #742 from longjon/direct-blob-buffers
Use Blob directly instead of shared_ptr for internal layer buffers
Jeff Donahue [Sun, 20 Jul 2014 20:13:01 +0000 (13:13 -0700)]
Merge pull request #743 from longjon/make-everything
Add an "everything" target to make for comprehensive build testing
Jonathan L Long [Sun, 20 Jul 2014 10:09:55 +0000 (03:09 -0700)]
Merge pull request #615 from kloudkl/top-k-argmax
Jonathan L Long [Sun, 20 Jul 2014 09:29:15 +0000 (02:29 -0700)]
remove unused includes in AccuracyLayer and ArgMaxLayer
Jonathan L Long [Sun, 20 Jul 2014 02:16:11 +0000 (19:16 -0700)]
Merge pull request #518 from longjon/fromto
Add From/To options for Forward and Backward
Jonathan L Long [Sun, 20 Jul 2014 01:51:15 +0000 (18:51 -0700)]
add an "everything" target to make for comprehensive build testing
The everything target builds all components, checks warnings, runs lint,
and runs tests.
Jonathan L Long [Sun, 20 Jul 2014 01:22:28 +0000 (18:22 -0700)]
use Blob directly instead of shared_ptr for WindowDataLayer buffers
Jonathan L Long [Sun, 20 Jul 2014 01:18:03 +0000 (18:18 -0700)]
use Blob directly instead of shared_ptr for ImageDataLayer buffers
Jonathan L Long [Sun, 20 Jul 2014 01:14:53 +0000 (18:14 -0700)]
use Blob directly instead of shared_ptr for DataLayer buffers
Jonathan L Long [Tue, 1 Jul 2014 19:52:48 +0000 (12:52 -0700)]
use Blob directly instead of shared_ptr for DropoutLayer::rand_vec_
This will simplify the implementation of Reshape by making allocation
automatic. There is no need for shared_ptr here, and this is more
uniform with buffers used by other layers.
Jonathan L Long [Tue, 1 Jul 2014 20:20:01 +0000 (13:20 -0700)]
use Blob directly instead of shared_ptr for PoolingLayer::max_idx_
Again, this is more uniform, and will simplify the implementation of
Reshape.
Jonathan L Long [Sun, 20 Jul 2014 01:02:33 +0000 (18:02 -0700)]
use Blob directly instead of shared_ptr for InnerProductLayer::bias_multiplier_
Jonathan L Long [Tue, 1 Jul 2014 06:11:11 +0000 (23:11 -0700)]
use Blob directly instead of shared_ptr for ConvolutionLayer::bias_multiplier_
This will make layer reshaping easier and more uniform, and is
consistent with col_buffer_.
Kai Li [Sat, 19 Jul 2014 16:42:12 +0000 (00:42 +0800)]
Fix style issues in accuracy & argmax layer
Kai Li [Sat, 12 Jul 2014 16:29:54 +0000 (00:29 +0800)]
Include <utility> for pair in the accuracy layer
Kai Li [Fri, 11 Jul 2014 02:22:01 +0000 (10:22 +0800)]
Limit the comparison functions to have file scope
Kai Li [Thu, 10 Jul 2014 01:38:03 +0000 (09:38 +0800)]
Add more test cases for the accuracy layer
Kai Li [Thu, 10 Jul 2014 01:37:34 +0000 (09:37 +0800)]
Refactor the accuracy layer with std::partial_sort
Kai Li [Thu, 10 Jul 2014 01:06:11 +0000 (09:06 +0800)]
Move compararing function from common_layers to argmax_layer
Kai Li [Wed, 9 Jul 2014 09:30:05 +0000 (17:30 +0800)]
Use std::partial_sort in the ArgMaxLayer as suggested by @shuokay
Kai Li [Tue, 8 Jul 2014 14:24:20 +0000 (22:24 +0800)]
Simplify the top-k argmax layer using std::sort
Kai Li [Sat, 5 Jul 2014 09:17:12 +0000 (17:17 +0800)]
Add the test cases for the mulitple top predictions argmax layer
Kai Li [Fri, 4 Jul 2014 09:19:24 +0000 (17:19 +0800)]
Extend the ArgMaxLayer to output top k results
Jeff Donahue [Fri, 18 Jul 2014 23:37:50 +0000 (16:37 -0700)]
Merge pull request #732 from longjon/travis-pycaffe
Have Travis build pycaffe
Jonathan L Long [Fri, 18 Jul 2014 22:48:36 +0000 (15:48 -0700)]
Travis builds pycaffe
Jonathan L Long [Fri, 18 Jul 2014 23:13:24 +0000 (16:13 -0700)]
use /usr instead of /usr/local for default Python include/lib
This makes the default settings usable with Travis/Ubuntu Python
Jonathan L Long [Fri, 18 Jul 2014 22:39:17 +0000 (15:39 -0700)]
test consistency of From/To Forward/Backward
Jonathan L Long [Fri, 18 Jul 2014 22:36:39 +0000 (15:36 -0700)]
comment in net.hpp to explain subtleties of From/To on DAGs
Jonathan L Long [Thu, 19 Jun 2014 04:03:28 +0000 (21:03 -0700)]
pycaffe: expose Forward/Backward From/To as kwargs start and end
Jonathan L Long [Tue, 18 Feb 2014 00:35:38 +0000 (16:35 -0800)]
add Net::Forward/Backward From/To
Yangqing Jia [Fri, 18 Jul 2014 22:18:14 +0000 (15:18 -0700)]
reapply namespace change
Yangqing Jia [Fri, 18 Jul 2014 20:45:41 +0000 (13:45 -0700)]
Merge pull request #730 from Yangqing/sweep
cosmetics: add syntax = proto2
Yangqing Jia [Fri, 18 Jul 2014 20:32:25 +0000 (13:32 -0700)]
cosmetics: add syntax = proto2
Yangqing Jia [Fri, 18 Jul 2014 00:53:24 +0000 (17:53 -0700)]
Merge pull request #718 from Yangqing/dev
More namespace cleaning.
Yangqing Jia [Fri, 18 Jul 2014 00:01:17 +0000 (17:01 -0700)]
lint
Jeff Donahue [Thu, 17 Jul 2014 19:46:18 +0000 (12:46 -0700)]
Merge pull request #716 from Yangqing/dev
Minor Code Cleaning
Yangqing Jia [Thu, 17 Jul 2014 19:18:27 +0000 (12:18 -0700)]
compute_image_mean namespace fix.
Evan Shelhamer [Thu, 17 Jul 2014 10:35:00 +0000 (12:35 +0200)]
Merge pull request #561 from shelhamer/cpu-only-build
CPU-only build
no GPU, no CUDA, no problem!
Evan Shelhamer [Thu, 17 Jul 2014 09:32:29 +0000 (11:32 +0200)]
check CPU-only everything, CPU + GPU build with travis
Evan Shelhamer [Wed, 16 Jul 2014 13:24:48 +0000 (15:24 +0200)]
relax benchmark test timing for cheap hardware / CI build
Evan Shelhamer [Thu, 17 Jul 2014 09:57:56 +0000 (11:57 +0200)]
switch travis build to CPU-only flag, drop runtestnogpu target
Evan Shelhamer [Tue, 15 Jul 2014 13:56:36 +0000 (15:56 +0200)]
collect CUDA includes and calls, separate from CPU-only mode, leave out
- collect CUDA includes in device_alternate.hpp
- add guards for CUDA code
- move GPU code into cu from cpp
- make CUDA includes and libraries conditional in Makefile
- drop CUDA dependency from travis-ci build
Evan Shelhamer [Mon, 30 Jun 2014 04:11:44 +0000 (21:11 -0700)]
add guards to drop GPU code in CPU-only mode
Evan Shelhamer [Mon, 30 Jun 2014 03:47:40 +0000 (20:47 -0700)]
stub out GPU layer methods to crash loudly in CPU-only mode
Evan Shelhamer [Mon, 30 Jun 2014 03:45:41 +0000 (20:45 -0700)]
configure Makefile for CPU-only build
Yangqing Jia [Wed, 16 Jul 2014 23:22:52 +0000 (16:22 -0700)]
neuron_layers.hpp should not need to include leveldb
Yangqing Jia [Wed, 16 Jul 2014 23:15:09 +0000 (16:15 -0700)]
move using statements inside namespace caffe to avoid polluting the whole name space.
Jeff Donahue [Tue, 15 Jul 2014 22:12:51 +0000 (15:12 -0700)]
Another bugfix related to my CPU/GPU test changes: make NetTest a
MultiDeviceTest (Caffe:set_mode(Caffe::CPU/GPU) isn't run without this).
Jeff Donahue [Tue, 15 Jul 2014 21:08:24 +0000 (14:08 -0700)]
Merge pull request #660 from jeffdonahue/param-propagate-down
Conditionally compute parameter gradients
Jeff Donahue [Tue, 15 Jul 2014 20:15:36 +0000 (13:15 -0700)]
Add Net Test to verify correct param_propagate_down behavior.
Jeff Donahue [Wed, 9 Jul 2014 22:34:44 +0000 (15:34 -0700)]
Use Blobs instead of SyncedMemorys for the bias_multiplier_'s.
Jeff Donahue [Wed, 9 Jul 2014 22:05:57 +0000 (15:05 -0700)]
Make ConvolutionLayer and InnerProductLayer abide by param_propagate_down_
Jeff Donahue [Wed, 9 Jul 2014 21:47:46 +0000 (14:47 -0700)]
Add param_propagate_down_ vector to layer, populate according to
blobs_lr in Net::Init
Jeff Donahue [Wed, 9 Jul 2014 21:25:55 +0000 (14:25 -0700)]
use layer_param instead of layers_[layer_id]->layer_param()
Jeff Donahue [Tue, 15 Jul 2014 20:47:14 +0000 (13:47 -0700)]
Fix Makefile warning error message.
Evan Shelhamer [Tue, 15 Jul 2014 14:06:59 +0000 (16:06 +0200)]
[example] elaborate net surgery description
Evan Shelhamer [Tue, 15 Jul 2014 08:20:10 +0000 (10:20 +0200)]
fix choice of clang++ in OS X build
Jeff Donahue [Tue, 15 Jul 2014 07:08:58 +0000 (00:08 -0700)]
Merge pull request #698 from jeffdonahue/softmax-layer-test-devices-fix
Fix SoftmaxLayerTest: forgot to change this one to use DtypesAndDevices;
Jeff Donahue [Tue, 15 Jul 2014 06:49:36 +0000 (23:49 -0700)]
Fix SoftmaxLayerTest: forgot to change this one to use DtypesAndDevices;
was causing Travis build to randomly fail if a previous test had set the
mode to GPU (which no test that is run by 'make runtestnogpu' should, so
I guess there's another bug somewhere).
Jeff Donahue [Tue, 15 Jul 2014 06:16:21 +0000 (23:16 -0700)]
Merge pull request #696 from rbgirshick/matlab_2014a_fix
Fix bug that breaks MATLAB 2014a compilation
Ross Girshick [Tue, 15 Jul 2014 00:04:43 +0000 (17:04 -0700)]
fix bug that breaks MATLAB 2014a compilation
Jeff Donahue [Mon, 14 Jul 2014 21:13:50 +0000 (14:13 -0700)]
Merge pull request #694 from jeffdonahue/travis-tweaks
Travis tweaks -- make --keep-going; build on all branches; replace CUSTOM_CXX <eom>
Jeff Donahue [Mon, 14 Jul 2014 20:46:09 +0000 (13:46 -0700)]
Add blank lines between fields for readability.
Jeff Donahue [Mon, 14 Jul 2014 20:43:08 +0000 (13:43 -0700)]
Replace CUSTOM_CXX env var to specify non-default C++ compiler.
Jeff Donahue [Mon, 14 Jul 2014 20:37:13 +0000 (13:37 -0700)]
Run Travis build on all branches (remove whitelist from .travis.yml).
Jeff Donahue [Mon, 14 Jul 2014 20:31:13 +0000 (13:31 -0700)]
Add --keep-going flag to first make so that any targets that can be
built are built; write out full -j flag as --jobs for clarity.
Evan Shelhamer [Mon, 14 Jul 2014 17:54:48 +0000 (19:54 +0200)]
fix Makefile comment about library names
Jeff Donahue [Mon, 14 Jul 2014 09:56:08 +0000 (02:56 -0700)]
Merge pull request #685 from jeffdonahue/seed-hinge-loss-test
Seed HingeLossLayerTest; bad values can cause test (and therefore Travis build) to fail
Jeff Donahue [Mon, 14 Jul 2014 09:44:28 +0000 (02:44 -0700)]
Seed HingeLossLayerTest; bad values can cause test (and therefore Travis
CI build) to fail
Evan Shelhamer [Mon, 14 Jul 2014 09:27:24 +0000 (11:27 +0200)]
Merge pull request #681 from BVLC/travisci
continuous integration for build, test, and lint by Travis CI
Jeff Donahue [Sun, 13 Jul 2014 05:37:10 +0000 (22:37 -0700)]
Travis build failure wasn't working for lint/warn because they didn't
exit with a non-zero code -- this fixes that.
Jeff Donahue [Sat, 12 Jul 2014 11:27:36 +0000 (04:27 -0700)]
@jeffdonahue's tweaks to .travis.yml config:
-Add (CPU-only) test, lint, warn and parallel (-j 4) to travis CI build.
-Add /usr/local/lib to travis config (seems needed for LMDB).
-Put export in "before_script"; disable clang build -- doesn't work on Linux.
-Cache Ubuntu apt packages.
-Install bc package to hopefully suppress "bc: not found" errors
-Get apt packages before_install as suggested by Travis official docs
-Remove specified email address and IRC notifications (emails are sent
to the committer by default; others can view build results in public
Travis feed, on pull requests, etc.).
Jeff Donahue [Sat, 12 Jul 2014 17:00:44 +0000 (10:00 -0700)]
-Override the default compiler by specifying a CXX in
Makefile.config instead of CUSTOM_CXX, as Travis exports CXX
as the compiler env variable name.
-Change TEST_HDFS -> TEST_HXX_SRCS.
huy [Thu, 10 Jul 2014 06:02:06 +0000 (23:02 -0700)]
@huyng's .travis.yml integration configuration file to install and test Caffe.
Jeff Donahue [Sat, 12 Jul 2014 13:58:35 +0000 (06:58 -0700)]
Move test headers to include/.
Add a test param to test both CPU and GPU (with both float and double
Dtypes).
Sergey Karayev [Sat, 12 Jul 2014 16:25:23 +0000 (09:25 -0700)]
back-merging [docs] changes and web demo [example] addition; updating
net_surgery example to new format
Conflicts:
docs/getting_pretrained_models.md
docs/index.md
Sergey Karayev [Sat, 12 Jul 2014 16:20:07 +0000 (09:20 -0700)]
Merge pull request #679 from sergeyk/master
[example] image classification web demo
Sergey Karayev [Sat, 12 Jul 2014 02:23:47 +0000 (19:23 -0700)]
[example] image classification web demo
Sergey Karayev [Sat, 12 Jul 2014 00:20:46 +0000 (17:20 -0700)]
Merge pull request #675 from sergeyk/master
Revamped docs system
Sergey Karayev [Sat, 12 Jul 2014 00:19:35 +0000 (17:19 -0700)]
[docs] shelhamer's minor suggestions
Sergey Karayev [Fri, 11 Jul 2014 21:00:56 +0000 (14:00 -0700)]
[docs] updated instructions for contributing documentation
Sergey Karayev [Fri, 11 Jul 2014 20:28:55 +0000 (13:28 -0700)]
[docs] moved example md’s to examples/**/md’s and added script to gather
them for publication
Sergey Karayev [Fri, 11 Jul 2014 20:27:38 +0000 (13:27 -0700)]
[docs] cosmetic
Evan Shelhamer [Fri, 11 Jul 2014 04:09:41 +0000 (06:09 +0200)]
fix link for caffe_rcnn_imagenet_model
Evan Shelhamer [Fri, 11 Jul 2014 04:07:55 +0000 (06:07 +0200)]
Back-merge documentation and fixes
* master:
host materials on dl.caffe.berkeleyvision.org
fix caffe acm-mm paper link
[docs] readme
[docs] got rid of redundant README, updated development instructions
[docs] reworked index page, got rid of publications page
point to reference performance from installation, add GTX 770
acknowledge BVLC PI Trevor Darrell for advising Caffe
add latest CUDA arch to fix invalid device function errors
switch language to "related publications"
add publication section to homepage
fix caffe paper link -- still hasn't appeared on arxiv yet
Added top-1 and top-5 accuracy for the caffe networks to docs
add skeleton of the Caffe publications page
Update docs on building boost on OSX for the python wrappers
fix OSX 10.9 homebrew CXX doc
caffe.Net preprocessing members belong to object, not class
10.9 install doc formatting
Evan Shelhamer [Fri, 11 Jul 2014 04:01:42 +0000 (06:01 +0200)]
host materials on dl.caffe.berkeleyvision.org
Evan Shelhamer [Thu, 10 Jul 2014 12:45:44 +0000 (14:45 +0200)]
fix caffe acm-mm paper link
longjon [Thu, 10 Jul 2014 00:18:33 +0000 (17:18 -0700)]
Merge pull request #633 from kloudkl/cpu-only-memcpy
Support CPU only memcpy
Kai Li [Wed, 9 Jul 2014 23:50:31 +0000 (07:50 +0800)]
Replace cudaMemcpy with caffe_gpu_memcpy in SyncedMemory per @longjon
Kai Li [Wed, 9 Jul 2014 17:57:16 +0000 (01:57 +0800)]
Implement @Yangqing's solution to copy memory in the SyncedMemory
Kai Li [Wed, 9 Jul 2014 02:12:07 +0000 (10:12 +0800)]
Switch to GPU mode when pointer is move to or from GPU in SyncedMemory
Kai Li [Wed, 9 Jul 2014 01:47:28 +0000 (09:47 +0800)]
Check the GPU mode to decide which memcpy to use
Kai Li [Mon, 7 Jul 2014 07:26:24 +0000 (15:26 +0800)]
Avoid using cudaMemcpy for memcpy when there is no GPU and CUDA driver