Przemysław Dolata [Tue, 21 Aug 2018 11:11:30 +0000 (13:11 +0200)]
Merge pull request #6455 from lengly/patch-1
Fix typo in inner_product_layer.cpp
Przemysław Dolata [Tue, 21 Aug 2018 07:44:25 +0000 (09:44 +0200)]
Merge pull request #6346 from jerryz123/HDF5_config
Added USE_HDF5 flag to build
Yuda Liu [Sun, 8 Jul 2018 07:46:43 +0000 (15:46 +0800)]
Update inner_product_layer.cpp
Przemysław Dolata [Mon, 20 Aug 2018 05:55:36 +0000 (07:55 +0200)]
Merge pull request #6320 from Noiredd/clip
Clip layer - resurrection
Jerry Zhao [Mon, 16 Apr 2018 09:26:08 +0000 (02:26 -0700)]
Cherry-picked USE_HDF5 from Android branch
Przemysław Dolata [Fri, 17 Aug 2018 12:45:04 +0000 (14:45 +0200)]
Clip layer documentation
Noiredd [Fri, 30 Mar 2018 09:43:44 +0000 (11:43 +0200)]
test case fix for Clip layer gradient
minor lint fixes
Przemysław Dolata [Fri, 17 Aug 2018 09:54:36 +0000 (11:54 +0200)]
Merge pull request #6282 from Noiredd/pooling-mode
PoolingLayer customizable output shape rounding mode
Harm Berntsen [Mon, 18 Jan 2016 10:41:14 +0000 (11:41 +0100)]
Add clip layer
Evan Shelhamer [Fri, 8 Jun 2018 03:57:31 +0000 (23:57 -0400)]
Merge pull request #6238 from mitar/manual-sgd
[pycaffe] expose interface for manual, step-by-step optimization
Noiredd [Wed, 7 Mar 2018 12:57:56 +0000 (13:57 +0100)]
PoolingLayer customizable output shape rounding mode
Valentin Tolmer [Wed, 22 Jun 2016 00:12:57 +0000 (17:12 -0700)]
[pycaffe] test solver update
Mitar [Sun, 10 Dec 2017 11:11:26 +0000 (03:11 -0800)]
[pycaffe] expose mutable solver parameter, base lr, and effective lr
`solver.lr` is the effective learning rate in use while `solver.base_lr`
is the configured learning rate at initialization. the solver parameter
is now editable for setting fields that are in use throughout the
lifetime of the solver, such as the maximum iteration.
Mitar [Sat, 28 Oct 2017 09:52:35 +0000 (02:52 -0700)]
increment iteration during update, not step
with update exposed it is important to increment the iteration when an
update is made, whether by step or update alone. more fundementally,
it's the update that defines an iterationa, so this is a natural place
for the increment.
Mitar [Sat, 28 Oct 2017 09:35:17 +0000 (02:35 -0700)]
[pycaffe] expose solver update to do manual solving
a sketch of `solver.step()` done out manually:
1. `solver.net.forward()`
2. `solver.net.backward()`
3. `solver.net.apply_update()`
4. `solver.net.clear_param_diffs()`
Evan Shelhamer [Wed, 6 Jun 2018 20:16:35 +0000 (16:16 -0400)]
Merge pull request #6390 from open-cv/fix-6389
drop unnecessary cast
fix issue #6389
Kuang Fangjun [Tue, 8 May 2018 02:46:54 +0000 (10:46 +0800)]
fix issue #6389
Evan Shelhamer [Wed, 6 Jun 2018 20:13:36 +0000 (16:13 -0400)]
Merge pull request #6388 from open-cv/fix-6387
fix missing syncedmem constness
fix issue #6387
Przemysław Dolata [Tue, 17 Apr 2018 12:44:25 +0000 (14:44 +0200)]
Merge pull request #6219 from uhfband/fix-caffe_pb2-module
Fix incorrect namespace for pycaffe submodule caffe_pb2
Kuang Fangjun [Tue, 8 May 2018 02:36:44 +0000 (10:36 +0800)]
fix issue #6387.
Przemysław Dolata [Thu, 12 Apr 2018 06:13:23 +0000 (08:13 +0200)]
Merge pull request #6342 from Noiredd/gauss-fill-test-fix
Gaussian filler tests adjustments
Przemysław Dolata [Wed, 11 Apr 2018 13:32:25 +0000 (15:32 +0200)]
tweaked Gaussian filler tests for less false fails
Przemysław Dolata [Tue, 10 Apr 2018 22:20:12 +0000 (00:20 +0200)]
Revised guidelines for GitHub issues (#6327)
revised guidelines for GitHub issues and caffe-users posts
Viktor Richter [Mon, 9 Apr 2018 14:51:04 +0000 (16:51 +0200)]
Fix cmake < v3.7 compatibility in Cuda.cmake (#6338)
Fix for compatibility with CMake <3.7 (related conversation under 37e4289)
* Fix cmake < v3.7 compatibility in Cuda.cmake
* Fix version test variable naming in Cuda.cmake
Przemysław Dolata [Mon, 9 Apr 2018 07:28:53 +0000 (09:28 +0200)]
Merge pull request #6336 from Coderx7/master
Minor correction concerning compilation compatibility with CUDA 9.0 (see #6237)
Seyyed Hossein Hasanpour [Mon, 9 Apr 2018 05:36:49 +0000 (10:06 +0430)]
Minor correction concerning compilation compatibility with CUDA 9.0
Since CUDA 9.0 doesn't support sm_20 and sm_21 anymore. This PR allows Caffe to compile with CUDA 9.0 and newer versions successfully.
This addresses the discussion concerning [#6237](https://github.com/BVLC/caffe/pull/6237) in [master (#6237)](https://github.com/BVLC/caffe/commit/
cb150eca6d593ddb85e53acd05b8dcf709ea8337#diff-
2004a3d3e6b4ed2e2812bb0b4b998f18)
Przemysław Dolata [Fri, 6 Apr 2018 14:16:48 +0000 (16:16 +0200)]
Merge pull request #6237 from jasjuang/master
Make caffe compile through cmake with CUDA 9.1
Przemysław Dolata [Tue, 27 Mar 2018 11:10:42 +0000 (13:10 +0200)]
Merge pull request #6277 from twmht/dev_issue_6268
check Embed index in debug mode
Mikhail Antonenka [Sat, 17 Mar 2018 15:26:40 +0000 (18:26 +0300)]
Added Swish layer (#6002)
* added swish layer (cpu)
* swish layer: added tests
* swish layer: optimized backpropogation
* swish layer: added cuda implementation
* swish layer: added beta parameter
* swish layer: incorporated sigmoid layer
* swish layer: fix comment of last added parameter
* swish layer: added REGISTER_LAYER_CLASS
Przemysław Dolata [Mon, 12 Mar 2018 09:52:04 +0000 (10:52 +0100)]
Merge pull request #6286 from Noiredd/bilinear-filler-fix
BilinearFiller tests refactored
Noiredd [Fri, 9 Mar 2018 12:14:32 +0000 (13:14 +0100)]
bilinear filter test refactor
Przemysław Dolata [Fri, 9 Mar 2018 08:19:21 +0000 (09:19 +0100)]
Merge pull request #6278 from Noiredd/filler-fix
1D blob handling in MSRA/Xavier fillers + new filler tests
Noiredd [Thu, 8 Mar 2018 13:02:10 +0000 (14:02 +0100)]
Filler testing overhaul
nic25 [Wed, 7 Mar 2018 11:40:52 +0000 (03:40 -0800)]
Add lr_mult label to the network graph in draw_net.py (#6273)
draw_net.py refactoring and optional LR visualization
* refactoring `get_layer_label`
rewrote the function body to make it more streamlined.
does not affect inputs and outputs
* optionally visualize LR when drawing the network
adds an option to `python/draw_net.py` that allows to visualize information
about the learning rate multiplier (if relevant) when drawing the network's
graph.
Noiredd [Tue, 6 Mar 2018 12:39:49 +0000 (13:39 +0100)]
1D blob handling in MSRA/Xavier fillers
twmht [Fri, 23 Mar 2018 14:21:06 +0000 (22:21 +0800)]
check embed index in debug mode
Przemysław Dolata [Mon, 26 Feb 2018 09:56:10 +0000 (10:56 +0100)]
Merge pull request #6249 from Noiredd/legacy-tools
Remove legacy tools
Noiredd [Fri, 23 Feb 2018 13:45:59 +0000 (14:45 +0100)]
Remove legacy tools
Przemysław Dolata [Wed, 21 Feb 2018 10:39:39 +0000 (11:39 +0100)]
Merge pull request #6104 from YaYaB/master_bvlc
Check solver gamma and stepsize to avoid unexplained core dump
Przemysław Dolata [Tue, 20 Feb 2018 13:03:14 +0000 (14:03 +0100)]
Merge pull request #5149 from yanchen036/master
fix extension extraction in matchExt
Przemysław Dolata [Tue, 20 Feb 2018 12:11:00 +0000 (13:11 +0100)]
Merge pull request #6176 from qu1j0t3/simplify-pip-cmd
Simplify pip requirements installation
Przemysław Dolata [Tue, 20 Feb 2018 07:31:01 +0000 (08:31 +0100)]
Merge pull request #6240 from knsong/master
Fix filler compatibility for ND convolution
knsong [Sat, 17 Feb 2018 07:56:32 +0000 (15:56 +0800)]
Fix compatibility for ND convolution
yanchen036 [Mon, 2 Jan 2017 05:33:49 +0000 (13:33 +0800)]
bug fix: ext should not include the '.'
jasjuang [Wed, 14 Feb 2018 00:57:18 +0000 (16:57 -0800)]
fix cuda 9.1 compilation
Przemysław Dolata [Mon, 12 Feb 2018 08:35:27 +0000 (09:35 +0100)]
Merge pull request #6123 from IlyaOvodov/master
"weights" added to solver parameters, "snapshot_prefix" field default initialization
iovodov [Sat, 13 Jan 2018 14:19:45 +0000 (17:19 +0300)]
replacement of snapshot_prefix parameter if it is empty or points to a directory. See issue #6110 proposed improvement No.2
iovodov [Mon, 18 Dec 2017 19:39:03 +0000 (22:39 +0300)]
Weight parameter in solver is used in caffe.exe
Loading weights is moved from caffe.exe to solver class, so new "weights" solver parameter is used not only from command line but when caffe is used as library (including python)
corrected formatting
fixed line length
more formatting corrected
Yegor Bedarev [Thu, 8 Feb 2018 11:05:21 +0000 (18:05 +0700)]
Fix incorrect namespace for pycaffe submodule caffe_pb2 generated by protobuf
Przemysław Dolata [Tue, 6 Feb 2018 07:46:25 +0000 (08:46 +0100)]
Merge pull request #5598 from ZoroDerVonCodier/patch-1
Update euclidean_loss_layer.hpp with corrected reference in comment
Przemysław Dolata [Sat, 3 Feb 2018 16:41:46 +0000 (17:41 +0100)]
Merge pull request #6210 from mzsanford/classfier_deprecation_warning
Update Classifier and Detector to avoid deprecation warning
Matt Sanford [Fri, 2 Feb 2018 21:23:23 +0000 (13:23 -0800)]
Update Classifier and Detector to avoid deprecation warning
Przemysław Dolata [Fri, 2 Feb 2018 15:47:46 +0000 (16:47 +0100)]
Merge pull request #5545 from brunobowden/shape_mismatch_checks
Shape mismatch CHECK logging improvements
Przemysław Dolata [Thu, 1 Feb 2018 13:43:44 +0000 (14:43 +0100)]
corrected description of set_transpose in io.py
Przemysław Dolata [Wed, 31 Jan 2018 08:10:34 +0000 (09:10 +0100)]
Merge pull request #5865 from cheshirekow/fix/caffe_rpath
Fix cmake RPATH
Evan Shelhamer [Mon, 29 Jan 2018 19:31:14 +0000 (11:31 -0800)]
Merge pull request #6202 from shelhamer/fix-scratch-bottom-diff
Clear Scratch Diffs to Prevent Contaminating Backward through Splits
Evan Shelhamer [Mon, 29 Jan 2018 00:58:39 +0000 (16:58 -0800)]
explain use of scratch diffs in comments
a few layers make use of otherwise unused diffs to accumulate results,
but unless the diffs are cleared in forward this contaminates the
gradients when these layers share a bottom and their backward is
skipped.
Evan Shelhamer [Mon, 29 Jan 2018 00:42:07 +0000 (16:42 -0800)]
clear scratch use of accuracy bottom diff
linziyi [Thu, 18 Jan 2018 19:07:45 +0000 (03:07 +0800)]
clear scratch use of loss bottom diffs
Evan Shelhamer [Mon, 29 Jan 2018 00:48:26 +0000 (16:48 -0800)]
Merge pull request #5924 from bowang/cudnn_deconv
speed-up: add cuDNN deconvolution layer and test
Evan Shelhamer [Sun, 28 Jan 2018 00:39:59 +0000 (16:39 -0800)]
Merge pull request #6201 from shelhamer/official-aws-ami
installation: switch pointer to official AWS AMI
Evan Shelhamer [Sun, 28 Jan 2018 00:37:06 +0000 (16:37 -0800)]
docs: switch to official AWS AMI
Toby Thain [Mon, 15 Jan 2018 18:16:05 +0000 (13:16 -0500)]
Simplify pip invocation.
Przemysław Dolata [Wed, 10 Jan 2018 19:30:43 +0000 (20:30 +0100)]
Merge pull request #6048 from sclarkson/master
Fix Makefile parallel builds missing protobuf header
Przemysław Dolata [Wed, 3 Jan 2018 09:21:03 +0000 (10:21 +0100)]
Merge pull request #6121 from xerus/typo
Cuda.cmake: Fix a typo in a comment
Pavel Grunt [Wed, 20 Dec 2017 12:07:47 +0000 (13:07 +0100)]
Cuda.cmake: Fix a typo in a comment
YaYaB [Tue, 12 Dec 2017 16:16:59 +0000 (16:16 +0000)]
Add check values of gamma and stepsize to avoid unexplained core dump
Przemysław Dolata [Wed, 29 Nov 2017 11:28:31 +0000 (12:28 +0100)]
Merge pull request #6084 from Noiredd/accuracy-count-fix
Added count==0 safeguard to CPU accuracy calculation
Noiredd [Wed, 29 Nov 2017 10:48:29 +0000 (11:48 +0100)]
Added count==0 safeguard to CPU accuracy calculation
Przemysław Dolata [Tue, 28 Nov 2017 10:36:47 +0000 (11:36 +0100)]
Merge pull request #6079 from Noiredd/cuda9-makefile
Makefile example comments for CUDA 9.0 compatibility
Noiredd [Tue, 28 Nov 2017 10:00:51 +0000 (11:00 +0100)]
Makefile example comments for CUDA 9.0 compatibility
sclarkson [Sun, 12 Nov 2017 02:20:11 +0000 (18:20 -0800)]
Fix Makefile parallel builds missing protobuf header
Przemysław Dolata [Tue, 7 Nov 2017 11:56:12 +0000 (12:56 +0100)]
Merge pull request #5972 from icyblade/icyblade-patch-1
add support for cuDNN v7
Przemysław Dolata [Mon, 6 Nov 2017 14:42:07 +0000 (15:42 +0100)]
Merge pull request #5719 from leemgs/upstream-issue5718
Fix: mean shape incompatible with input shape
Geunsik Lim [Thu, 29 Jun 2017 04:12:53 +0000 (13:12 +0900)]
Fix: mean shape in compatible with input shape
This commit is to fix issue #5718.
* reference:
1. https://groups.google.com/forum/#!topic/caffe-users/nBpWJCcJoCU
2. https://stackoverflow.com/questions/
28692209/using-gpu-despite-setting-cpu-only-yielding-unexpected-keyword-argument
Signed-off-by: Geunsik Lim <geunsik.lim@samsung.com>
Przemysław Dolata [Thu, 2 Nov 2017 11:52:30 +0000 (12:52 +0100)]
Merge pull request #5813 from jqueguiner/patch-1
fix web demo install instruction link
Przemysław Dolata [Thu, 2 Nov 2017 11:46:05 +0000 (12:46 +0100)]
Merge pull request #5704 from ArneSuppe/dupDistDirFix
Fixed bug where make distribute duplicates python files
Przemysław Dolata [Thu, 2 Nov 2017 11:34:54 +0000 (12:34 +0100)]
Merge pull request #5969 from developius/fix-default-mode-warning
Fix default mode warning in io.resize_image
Przemysław Dolata [Thu, 2 Nov 2017 10:56:53 +0000 (11:56 +0100)]
Merge pull request #5866 from cijianzy/update_link_to_google_style_guide
Update link to google style guide.
Przemysław Dolata [Thu, 19 Oct 2017 07:53:59 +0000 (09:53 +0200)]
Merge pull request #5973 from Noiredd/pytest
Add absolute tolerance to test_net.py to prevent random Travis fails
Mikhail Antonenka [Tue, 17 Oct 2017 15:18:32 +0000 (18:18 +0300)]
infogain loss: fix bottom blobs description
Evan Shelhamer [Wed, 11 Oct 2017 17:02:53 +0000 (10:02 -0700)]
Merge pull request #5925 from BVLC/williford-install-ubuntu-16.04-patch-1
[docs] packages needed by Ubuntu 16.04, not just Ubuntu 14.04
Noiredd [Wed, 11 Oct 2017 09:04:18 +0000 (11:04 +0200)]
Add absolute tolerance to test_net.py to prevent random Travis fails
Icyblade Dai [Wed, 11 Oct 2017 07:29:21 +0000 (15:29 +0800)]
add supports for cuDNN v7
Shai [Thu, 10 Aug 2017 07:07:19 +0000 (10:07 +0300)]
upgrading Accuracy layer: (1) efficient CPU implementation O(L) for top_k, no need for fancy priority_queue etc. (2) GPU implementation
Finnian Anderson [Tue, 10 Oct 2017 11:15:13 +0000 (12:15 +0100)]
Fix default mode warning in io.resize_image
Signed-off-by: Finnian Anderson <get@finnian.io>
Evan Shelhamer [Tue, 3 Oct 2017 21:20:13 +0000 (14:20 -0700)]
[docs] fix link to `AbsVal` layer
Evan Shelhamer [Tue, 3 Oct 2017 02:05:50 +0000 (03:05 +0100)]
Merge pull request #5713 from Noiredd/filler
fix bilinear filler (and make constant filler more strict, as it should be)
Noiredd [Mon, 2 Oct 2017 12:39:31 +0000 (14:39 +0200)]
Fixed bilinear filler, added tests
Evan Shelhamer [Tue, 19 Sep 2017 11:32:56 +0000 (12:32 +0100)]
Merge pull request #5904 from longjon/gpu-ptr
Expose GPU pointers to Python
Jonathan R. Williford [Fri, 15 Sep 2017 09:38:16 +0000 (11:38 +0200)]
Packages needed by Ubuntu 16.04 also
This line is needed for Ubuntu 16.04:
sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev
For reference:
* https://github.com/BVLC/caffe/wiki/Ubuntu-16.04-or-15.10-Installation-Guide
* https://youtu.be/DnIs4DRjNL4
Bo Wang [Thu, 14 Sep 2017 22:03:21 +0000 (15:03 -0700)]
Fix format
Bo Wang [Sun, 10 Sep 2017 04:57:20 +0000 (21:57 -0700)]
Implement CuDNN-based deconvolution layer and test
Takuya Narihira [Wed, 11 Mar 2015 23:23:39 +0000 (16:23 -0700)]
Expose GPU pointers to Python
The pointers could be used by CUDA wrapper libraries in Python such as
PyCUDA, gnumpy, Theano etc.
Jon Long [Wed, 6 Sep 2017 23:09:16 +0000 (16:09 -0700)]
Merge pull request #5900 from wasnot/fix/py3-division-compat
Fix division operator for Compatibility of python 3 in classifier.py
wasnot [Wed, 6 Sep 2017 03:32:39 +0000 (12:32 +0900)]
modified division operator for compatibility of python 3
此间 [Mon, 21 Aug 2017 23:36:09 +0000 (07:36 +0800)]
Update link to google style guide.
Josh Bialkowski [Mon, 21 Aug 2017 15:31:25 +0000 (08:31 -0700)]
Fix caffe rpath
The logic for setting the library RPATH checks whether or not
${CMAKE_INSTALL_PREFIX}/lib is a system directory, and if not
adds it to the library RPATH. However, caffe does not install
to ${CMAKE_INSTALL_PREFIX}/lib, it installs to
${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_LIBDIR} (from
GNUInstallDirs). CMAKE_INSTALL_LIBDIR may be something like
"lib/x86_64-linux-gnu"
Cyprien Noel [Thu, 17 Aug 2017 01:24:32 +0000 (18:24 -0700)]
Update README.md
Mention SKX support
Evan Shelhamer [Sun, 6 Aug 2017 11:11:29 +0000 (12:11 +0100)]
Merge pull request #5824 from ek9852/master
[cmake] fix hardcode xcode path