platform/upstream/caffeonacl.git
9 years agoMerge pull request #806 from jeffdonahue/travis-minimal-cuda
Jeff Donahue [Mon, 28 Jul 2014 04:10:31 +0000 (21:10 -0700)]
Merge pull request #806 from jeffdonahue/travis-minimal-cuda

speed up Travis build by installing only the CUDA subpackages necessary for building

9 years agoOnly install the minimal CUDA subpackages necessary for building.
Jeff Donahue [Mon, 28 Jul 2014 02:02:58 +0000 (19:02 -0700)]
Only install the minimal CUDA subpackages necessary for building.

9 years agoMerge pull request #802 from jeffdonahue/caffe-bin-symlink
Yangqing Jia [Sun, 27 Jul 2014 20:04:33 +0000 (13:04 -0700)]
Merge pull request #802 from jeffdonahue/caffe-bin-symlink

Symlink to tool bins without the .bin extension

9 years agoSymlink to tool bins without the .bin extension
Jeff Donahue [Sun, 27 Jul 2014 19:50:13 +0000 (12:50 -0700)]
Symlink to tool bins without the .bin extension

9 years agoMerge pull request #800 from jeffdonahue/zero-indexed-train-iter
Jeff Donahue [Sun, 27 Jul 2014 19:13:03 +0000 (12:13 -0700)]
Merge pull request #800 from jeffdonahue/zero-indexed-train-iter

Make training iterations 0-indexed to display training loss in 0th iter

9 years agoMove loss display before lr display in ComputeUpdateValue.
Jeff Donahue [Sun, 27 Jul 2014 03:57:07 +0000 (20:57 -0700)]
Move loss display before lr display in ComputeUpdateValue.

9 years agoMake training iterations 0-indexed.
Jeff Donahue [Sun, 27 Jul 2014 00:03:27 +0000 (17:03 -0700)]
Make training iterations 0-indexed.

9 years agoFix my nonsensical variable names.
Jeff Donahue [Sun, 27 Jul 2014 02:53:15 +0000 (19:53 -0700)]
Fix my nonsensical variable names.

9 years agoMerge pull request #796 from jeffdonahue/solver-debug-info
Jeff Donahue [Sat, 26 Jul 2014 23:37:06 +0000 (16:37 -0700)]
Merge pull request #796 from jeffdonahue/solver-debug-info

Print blob debug info during training if SolverParameter "debug_info" field is set

9 years agoPrint just the mean absolute value (no sum/l1norm)
Jeff Donahue [Sat, 26 Jul 2014 22:18:41 +0000 (15:18 -0700)]
Print just the mean absolute value (no sum/l1norm)

9 years agoPrint blob L1 norms during forward/backward passes and updates if
Jeff Donahue [Sat, 26 Jul 2014 03:40:07 +0000 (20:40 -0700)]
Print blob L1 norms during forward/backward passes and updates if
new "debug_info" field in SolverParameter is set.

9 years agoMerge pull request #789 from Yangqing/sweep
Yangqing Jia [Sat, 26 Jul 2014 15:47:43 +0000 (08:47 -0700)]
Merge pull request #789 from Yangqing/sweep

Consolidate tools.

9 years agoLOG(ERROR)->LOG(FATAL), and misc script changes.
Yangqing Jia [Sat, 26 Jul 2014 15:37:05 +0000 (08:37 -0700)]
LOG(ERROR)->LOG(FATAL), and misc script changes.

9 years agoMerge pull request #793 from jeffdonahue/infogain-test
Jeff Donahue [Sat, 26 Jul 2014 02:18:47 +0000 (19:18 -0700)]
Merge pull request #793 from jeffdonahue/infogain-test

Add gradient checks for InfogainLossLayer

9 years agousing caffe::string to be consistent with other string definition.
Yangqing Jia [Sat, 26 Jul 2014 00:58:35 +0000 (17:58 -0700)]
using caffe::string to be consistent with other string definition.

9 years agoAdd gradient checks for infogain loss layer, letting it take the
Jeff Donahue [Thu, 24 Jul 2014 02:22:11 +0000 (19:22 -0700)]
Add gradient checks for infogain loss layer, letting it take the
infogain matrix as the third bottom blob.

9 years agooops, wrong deprecation message (should have --)
Yangqing Jia [Fri, 25 Jul 2014 17:11:26 +0000 (10:11 -0700)]
oops, wrong deprecation message (should have --)

9 years agolint
Yangqing Jia [Fri, 25 Jul 2014 16:52:30 +0000 (09:52 -0700)]
lint

9 years agoAdding new caffe binary that does everything in one binary; deprecating device_query...
Yangqing Jia [Fri, 25 Jul 2014 16:02:37 +0000 (09:02 -0700)]
Adding new caffe binary that does everything in one binary; deprecating device_query, finetune_net, net_speed_benchmark, train_net

9 years agoinclude benchmark.hpp
Yangqing Jia [Fri, 25 Jul 2014 16:02:03 +0000 (09:02 -0700)]
include benchmark.hpp

9 years agoMerge pull request #776 from Yangqing/sweep
Yangqing Jia [Thu, 24 Jul 2014 21:15:45 +0000 (14:15 -0700)]
Merge pull request #776 from Yangqing/sweep

Gflags dependency & misc

9 years agotravis: gflags still needs -fPIC, otherwise it makes caffe cry.
Yangqing Jia [Thu, 24 Jul 2014 18:05:58 +0000 (11:05 -0700)]
travis: gflags still needs -fPIC, otherwise it makes caffe cry.

9 years agoI am really bad at debugging travis.
Yangqing Jia [Thu, 24 Jul 2014 17:53:53 +0000 (10:53 -0700)]
I am really bad at debugging travis.

9 years agoturns out that glog does not compile with gflags 2.1 - compiling glog first and then...
Yangqing Jia [Thu, 24 Jul 2014 17:48:04 +0000 (10:48 -0700)]
turns out that glog does not compile with gflags 2.1 - compiling glog first and then gflags.

9 years agotravis - does adding cflags help? want to keep minimal invasion into gflags installation.
Yangqing Jia [Thu, 24 Jul 2014 17:28:06 +0000 (10:28 -0700)]
travis - does adding cflags help? want to keep minimal invasion into gflags installation.

9 years agotravis
Yangqing Jia [Thu, 24 Jul 2014 16:28:46 +0000 (09:28 -0700)]
travis

9 years agogflags should preferrably be installed before glog
Yangqing Jia [Thu, 24 Jul 2014 05:49:16 +0000 (22:49 -0700)]
gflags should preferrably be installed before glog

9 years agoadd gflags dependency doc
Yangqing Jia [Thu, 24 Jul 2014 05:47:58 +0000 (22:47 -0700)]
add gflags dependency doc

9 years agoadd gflags dependency to caffe.
Yangqing Jia [Thu, 24 Jul 2014 05:47:25 +0000 (22:47 -0700)]
add gflags dependency to caffe.

9 years agoMerge pull request #773 from qipeng/lrelu
Jeff Donahue [Wed, 23 Jul 2014 19:22:09 +0000 (12:22 -0700)]
Merge pull request #773 from qipeng/lrelu

Leaky ReLU numerical stability improvement

9 years agonumerical stability improvement
qipeng [Wed, 23 Jul 2014 18:48:19 +0000 (11:48 -0700)]
numerical stability improvement

9 years agoMerge pull request #766 from rbgirshick/matcaffe-cxx
Jeff Donahue [Wed, 23 Jul 2014 18:36:40 +0000 (11:36 -0700)]
Merge pull request #766 from rbgirshick/matcaffe-cxx

Use the same CXX for matcaffe as for the rest of the build

10 years agoUse the same CXX for matcaffe as for the rest of the build
Ross Girshick [Wed, 23 Jul 2014 03:06:48 +0000 (20:06 -0700)]
Use the same CXX for matcaffe as for the rest of the build

10 years agoMerge pull request #762 from rbgirshick/upgrade-pad-pooling-layers
Jeff Donahue [Tue, 22 Jul 2014 20:12:18 +0000 (13:12 -0700)]
Merge pull request #762 from rbgirshick/upgrade-pad-pooling-layers

Correctly apply padding to pooling layers when upgrading from V0

10 years agoMerge pull request #760 from bhack/fix_loopcopy
Jeff Donahue [Tue, 22 Jul 2014 20:00:45 +0000 (13:00 -0700)]
Merge pull request #760 from bhack/fix_loopcopy

Fix repeated field copy loop

10 years agoCorrectly apply padding to pooling layers when upgrading from V0
Ross Girshick [Tue, 22 Jul 2014 19:51:48 +0000 (12:51 -0700)]
Correctly apply padding to pooling layers when upgrading from V0

10 years agoFix repeated field copy loop
bhack [Tue, 22 Jul 2014 17:25:45 +0000 (19:25 +0200)]
Fix repeated field copy loop

Fix lint algorithm include

Fix spaces

10 years agoMerge pull request #755 from jeffdonahue/bhack-split_dim
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)

10 years agoSliceLayer: post-rebase fixes, cleanup, etc. (some from changes
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.

10 years agoAdd split dim layer
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

10 years agoMerge pull request #740 from qipeng/lrelu
Jeff Donahue [Tue, 22 Jul 2014 04:51:16 +0000 (21:51 -0700)]
Merge pull request #740 from qipeng/lrelu

Leaky ReLU

10 years agoreduced multiplications & fixed unit test
qipeng [Sun, 20 Jul 2014 00:28:52 +0000 (17:28 -0700)]
reduced multiplications & fixed unit test

10 years agoleaky relu + unit test
qipeng [Sat, 19 Jul 2014 20:24:01 +0000 (13:24 -0700)]
leaky relu + unit test

10 years agoMerge pull request #752 from Yangqing/sweep
Jeff Donahue [Mon, 21 Jul 2014 17:25:59 +0000 (10:25 -0700)]
Merge pull request #752 from Yangqing/sweep

reapply namespace change

10 years agoMerge pull request #742 from longjon/direct-blob-buffers
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

10 years agoMerge pull request #743 from longjon/make-everything
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

10 years agoMerge pull request #615 from kloudkl/top-k-argmax
Jonathan L Long [Sun, 20 Jul 2014 10:09:55 +0000 (03:09 -0700)]
Merge pull request #615 from kloudkl/top-k-argmax

10 years agoremove unused includes in AccuracyLayer and ArgMaxLayer
Jonathan L Long [Sun, 20 Jul 2014 09:29:15 +0000 (02:29 -0700)]
remove unused includes in AccuracyLayer and ArgMaxLayer

10 years agoMerge pull request #518 from longjon/fromto
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

10 years agoadd an "everything" target to make for comprehensive build testing
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.

10 years agouse Blob directly instead of shared_ptr for WindowDataLayer buffers
Jonathan L Long [Sun, 20 Jul 2014 01:22:28 +0000 (18:22 -0700)]
use Blob directly instead of shared_ptr for WindowDataLayer buffers

10 years agouse Blob directly instead of shared_ptr for ImageDataLayer 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

10 years agouse Blob directly instead of shared_ptr for DataLayer 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

10 years agouse Blob directly instead of shared_ptr for DropoutLayer::rand_vec_
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.

10 years agouse Blob directly instead of shared_ptr for PoolingLayer::max_idx_
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.

10 years agouse Blob directly instead of shared_ptr for InnerProductLayer::bias_multiplier_
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_

10 years agouse Blob directly instead of shared_ptr for ConvolutionLayer::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_.

10 years agoFix style issues in accuracy & argmax layer
Kai Li [Sat, 19 Jul 2014 16:42:12 +0000 (00:42 +0800)]
Fix style issues in accuracy & argmax layer

10 years agoInclude <utility> for pair in the accuracy layer
Kai Li [Sat, 12 Jul 2014 16:29:54 +0000 (00:29 +0800)]
Include <utility> for pair in the accuracy layer

10 years agoLimit the comparison functions to have file scope
Kai Li [Fri, 11 Jul 2014 02:22:01 +0000 (10:22 +0800)]
Limit the comparison functions to have file scope

10 years agoAdd more test cases for the accuracy layer
Kai Li [Thu, 10 Jul 2014 01:38:03 +0000 (09:38 +0800)]
Add more test cases for the accuracy layer

10 years agoRefactor the accuracy layer with std::partial_sort
Kai Li [Thu, 10 Jul 2014 01:37:34 +0000 (09:37 +0800)]
Refactor the accuracy layer with std::partial_sort

10 years agoMove compararing function from common_layers to argmax_layer
Kai Li [Thu, 10 Jul 2014 01:06:11 +0000 (09:06 +0800)]
Move compararing function from common_layers to argmax_layer

10 years agoUse std::partial_sort in the ArgMaxLayer as suggested by @shuokay
Kai Li [Wed, 9 Jul 2014 09:30:05 +0000 (17:30 +0800)]
Use std::partial_sort in the ArgMaxLayer as suggested by @shuokay

10 years agoSimplify the top-k argmax layer using std::sort
Kai Li [Tue, 8 Jul 2014 14:24:20 +0000 (22:24 +0800)]
Simplify the top-k argmax layer using std::sort

10 years agoAdd the test cases for the mulitple top predictions argmax layer
Kai Li [Sat, 5 Jul 2014 09:17:12 +0000 (17:17 +0800)]
Add the test cases for the mulitple top predictions argmax layer

10 years agoExtend the ArgMaxLayer to output top k results
Kai Li [Fri, 4 Jul 2014 09:19:24 +0000 (17:19 +0800)]
Extend the ArgMaxLayer to output top k results

10 years agoMerge pull request #732 from longjon/travis-pycaffe
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

10 years agoTravis builds pycaffe
Jonathan L Long [Fri, 18 Jul 2014 22:48:36 +0000 (15:48 -0700)]
Travis builds pycaffe

10 years agouse /usr instead of /usr/local for default Python include/lib
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

10 years agotest consistency of From/To Forward/Backward
Jonathan L Long [Fri, 18 Jul 2014 22:39:17 +0000 (15:39 -0700)]
test consistency of From/To Forward/Backward

10 years agocomment in net.hpp to explain subtleties of From/To on DAGs
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

10 years agopycaffe: expose Forward/Backward From/To as kwargs start and end
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

10 years agoadd Net::Forward/Backward From/To
Jonathan L Long [Tue, 18 Feb 2014 00:35:38 +0000 (16:35 -0800)]
add Net::Forward/Backward From/To

10 years agoreapply namespace change
Yangqing Jia [Fri, 18 Jul 2014 22:18:14 +0000 (15:18 -0700)]
reapply namespace change

10 years agoMerge pull request #730 from Yangqing/sweep
Yangqing Jia [Fri, 18 Jul 2014 20:45:41 +0000 (13:45 -0700)]
Merge pull request #730 from Yangqing/sweep

cosmetics: add syntax = proto2

10 years agocosmetics: add syntax = proto2
Yangqing Jia [Fri, 18 Jul 2014 20:32:25 +0000 (13:32 -0700)]
cosmetics: add syntax = proto2

10 years agoMerge pull request #718 from Yangqing/dev
Yangqing Jia [Fri, 18 Jul 2014 00:53:24 +0000 (17:53 -0700)]
Merge pull request #718 from Yangqing/dev

More namespace cleaning.

10 years agolint
Yangqing Jia [Fri, 18 Jul 2014 00:01:17 +0000 (17:01 -0700)]
lint

10 years agoMerge pull request #716 from Yangqing/dev
Jeff Donahue [Thu, 17 Jul 2014 19:46:18 +0000 (12:46 -0700)]
Merge pull request #716 from Yangqing/dev

Minor Code Cleaning

10 years agocompute_image_mean namespace fix.
Yangqing Jia [Thu, 17 Jul 2014 19:18:27 +0000 (12:18 -0700)]
compute_image_mean namespace fix.

10 years agoMerge pull request #561 from shelhamer/cpu-only-build
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!

10 years agocheck CPU-only everything, CPU + GPU build with travis
Evan Shelhamer [Thu, 17 Jul 2014 09:32:29 +0000 (11:32 +0200)]
check CPU-only everything, CPU + GPU build with travis

10 years agorelax benchmark test timing for cheap hardware / CI build
Evan Shelhamer [Wed, 16 Jul 2014 13:24:48 +0000 (15:24 +0200)]
relax benchmark test timing for cheap hardware / CI build

10 years agoswitch travis build to CPU-only flag, drop runtestnogpu target
Evan Shelhamer [Thu, 17 Jul 2014 09:57:56 +0000 (11:57 +0200)]
switch travis build to CPU-only flag, drop runtestnogpu target

10 years agocollect CUDA includes and calls, separate from CPU-only mode, leave out
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

10 years agoadd guards to drop GPU code in CPU-only mode
Evan Shelhamer [Mon, 30 Jun 2014 04:11:44 +0000 (21:11 -0700)]
add guards to drop GPU code in CPU-only mode

10 years agostub out GPU layer methods to crash loudly 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

10 years agoconfigure Makefile for CPU-only build
Evan Shelhamer [Mon, 30 Jun 2014 03:45:41 +0000 (20:45 -0700)]
configure Makefile for CPU-only build

10 years agoneuron_layers.hpp should not need to include leveldb
Yangqing Jia [Wed, 16 Jul 2014 23:22:52 +0000 (16:22 -0700)]
neuron_layers.hpp should not need to include leveldb

10 years agomove using statements inside namespace caffe to avoid polluting the whole name space.
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.

10 years agoAnother bugfix related to my CPU/GPU test changes: make NetTest a
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).

10 years agoMerge pull request #660 from jeffdonahue/param-propagate-down
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

10 years agoAdd Net Test to verify correct param_propagate_down behavior.
Jeff Donahue [Tue, 15 Jul 2014 20:15:36 +0000 (13:15 -0700)]
Add Net Test to verify correct param_propagate_down behavior.

10 years agoUse Blobs instead of SyncedMemorys for the bias_multiplier_'s.
Jeff Donahue [Wed, 9 Jul 2014 22:34:44 +0000 (15:34 -0700)]
Use Blobs instead of SyncedMemorys for the bias_multiplier_'s.

10 years agoMake ConvolutionLayer and InnerProductLayer abide by param_propagate_down_
Jeff Donahue [Wed, 9 Jul 2014 22:05:57 +0000 (15:05 -0700)]
Make ConvolutionLayer and InnerProductLayer abide by param_propagate_down_

10 years agoAdd param_propagate_down_ vector to layer, populate according to
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

10 years agouse layer_param instead of layers_[layer_id]->layer_param()
Jeff Donahue [Wed, 9 Jul 2014 21:25:55 +0000 (14:25 -0700)]
use layer_param instead of layers_[layer_id]->layer_param()

10 years agoFix Makefile warning error message.
Jeff Donahue [Tue, 15 Jul 2014 20:47:14 +0000 (13:47 -0700)]
Fix Makefile warning error message.

10 years ago[example] elaborate net surgery description
Evan Shelhamer [Tue, 15 Jul 2014 14:06:59 +0000 (16:06 +0200)]
[example] elaborate net surgery description