platform/upstream/caffe.git
9 years ago[examples] add Euclidean loss PythonLayer
Evan Shelhamer [Tue, 30 Jun 2015 22:35:21 +0000 (15:35 -0700)]
[examples] add Euclidean loss PythonLayer

9 years ago[examples] sequence and revise notebooks
Evan Shelhamer [Wed, 18 Mar 2015 02:01:02 +0000 (19:01 -0700)]
[examples] sequence and revise notebooks

- combine classification + filter visualization
- order by classification, learning LeNet, brewing logreg, and
  fine-tuning to flickr style
- improve flow of content in classification + filter visualization
- include solver needed for learning LeNet
- edit notebook descriptions for site catalogue

9 years ago[examples] flickr fine-tuning notebook
Yangqing Jia [Tue, 17 Mar 2015 17:50:42 +0000 (10:50 -0700)]
[examples] flickr fine-tuning notebook

9 years ago[examples] draft Python solving example
Jonathan L Long [Thu, 12 Mar 2015 08:03:51 +0000 (01:03 -0700)]
[examples] draft Python solving example

9 years agoMerge pull request #2086 from longjon/python-net-spec
Evan Shelhamer [Tue, 30 Jun 2015 20:33:54 +0000 (13:33 -0700)]
Merge pull request #2086 from longjon/python-net-spec

Python net specification

9 years ago[pytest] minimal testing of net specification
Jonathan L Long [Thu, 18 Jun 2015 20:25:08 +0000 (13:25 -0700)]
[pytest] minimal testing of net specification

9 years ago[examples] caffenet python spec
Jonathan L Long [Tue, 10 Mar 2015 00:43:53 +0000 (17:43 -0700)]
[examples] caffenet python spec

9 years ago[pycaffe] basic net specification
Jonathan L Long [Thu, 15 Jan 2015 09:27:42 +0000 (01:27 -0800)]
[pycaffe] basic net specification

9 years agoMerge pull request #2611 from dpaiton/test_net_bugfix
Evan Shelhamer [Tue, 23 Jun 2015 21:48:24 +0000 (14:48 -0700)]
Merge pull request #2611 from dpaiton/test_net_bugfix

[compatibility] fix syntax for old protobuf

9 years agocopyright 2015
Evan Shelhamer [Tue, 23 Jun 2015 16:59:22 +0000 (09:59 -0700)]
copyright 2015

9 years agoMerge pull request #2579 from lukeyeager/draw-fix
Evan Shelhamer [Tue, 23 Jun 2015 16:54:26 +0000 (09:54 -0700)]
Merge pull request #2579 from lukeyeager/draw-fix

small platform-specific bugfix for draw.py

9 years agoMerge pull request #2636 from berleon/master
Sergio Guadarrama [Mon, 22 Jun 2015 15:36:33 +0000 (08:36 -0700)]
Merge pull request #2636 from berleon/master

fixed _force_color check, fixes #2635

9 years agofixed _force_color check, fixes #2635
berleon [Mon, 22 Jun 2015 12:41:59 +0000 (14:41 +0200)]
fixed _force_color check, fixes #2635

9 years agoMerge pull request #2609 from lukeyeager/doc-fixes
Evan Shelhamer [Fri, 19 Jun 2015 22:39:59 +0000 (15:39 -0700)]
Merge pull request #2609 from lukeyeager/doc-fixes

[docs] fix "the life of a branch" formatting

9 years agofixed two bugs with prototext format
Dylan Paiton [Wed, 17 Jun 2015 00:57:07 +0000 (17:57 -0700)]
fixed two bugs with prototext format

The first bug was in InitUnsharedWeightsNet. Proto var was of type
ostringstream, which converted the bool bias_term into an int. I
wrote an inline conditional to convert the term into a string.
This allows backwards compatibility with earlier prototext
versions (e.g. version 2.3.0 on Redhat was failing without this).

The second bug was in the syntax for repeated bool parameters,
assigned to the propagate_down parameter. The style used for e.g.
propagate_down: [true,true] does not work with earlier prototext
versions (failed with version 2.3.0 on Redhat). New syntax works
for all versions.

9 years agoFix HTML display for docs
Luke Yeager [Tue, 16 Jun 2015 16:12:32 +0000 (09:12 -0700)]
Fix HTML display for docs

Doesn't change display in Markdown

9 years agoMerge pull request #2605 from ajschumacher/patch-2
Jeff Donahue [Tue, 16 Jun 2015 00:21:34 +0000 (17:21 -0700)]
Merge pull request #2605 from ajschumacher/patch-2

typo: "a fixed steps" to "at fixed steps"

9 years agotypo: "a fixed steps" to "at fixed steps"
Aaron Schumacher [Mon, 15 Jun 2015 23:53:05 +0000 (19:53 -0400)]
typo: "a fixed steps" to "at fixed steps"

fixing in the correct place as per @shelhamer's advice from #2602

9 years agoMerge pull request #2278 from sguada/db_split
Sergio Guadarrama [Mon, 15 Jun 2015 20:59:02 +0000 (13:59 -0700)]
Merge pull request #2278 from sguada/db_split

Db split

9 years ago[docs] drop out-of-date reference to dev branch
Evan Shelhamer [Wed, 10 Jun 2015 21:43:30 +0000 (17:43 -0400)]
[docs] drop out-of-date reference to dev branch

9 years agoSmall platform-specific bugfix for draw.py
Luke Yeager [Tue, 9 Jun 2015 17:45:16 +0000 (10:45 -0700)]
Small platform-specific bugfix for draw.py

Close #2376

On Gentoo and CentOS (and others?), you get this error:
    Warning: /tmp/tmpjqPQBC:6: string ran past end of line

9 years agoSplit db.hpp into leveldb_db.hpp and lmdb_db.hpp
Sergio Guadarrama [Tue, 31 Mar 2015 00:29:50 +0000 (17:29 -0700)]
Split db.hpp into leveldb_db.hpp and lmdb_db.hpp

9 years agoMerge pull request #2551 from ShaggO/loglayer
Evan Shelhamer [Thu, 4 Jun 2015 19:14:43 +0000 (12:14 -0700)]
Merge pull request #2551 from ShaggO/loglayer

LogLayer gpu functionality moved to .cu file

9 years agoLogLayer gpu functionality moved to .cu file
Malte Stær Nissen [Thu, 4 Jun 2015 09:13:45 +0000 (11:13 +0200)]
LogLayer gpu functionality moved to .cu file

9 years ago[bug] fix double instantiation of GPU methods in LogLayer
Evan Shelhamer [Wed, 3 Jun 2015 17:46:53 +0000 (10:46 -0700)]
[bug] fix double instantiation of GPU methods in LogLayer

9 years agoMerge pull request #2089 from jeffdonahue/reduction-layer
Jeff Donahue [Wed, 3 Jun 2015 04:24:02 +0000 (21:24 -0700)]
Merge pull request #2089 from jeffdonahue/reduction-layer

ReductionLayer

9 years agoAdd ReductionLayer to reduce any number of "tail" axes to a scalar value
Jeff Donahue [Mon, 3 Nov 2014 01:21:37 +0000 (17:21 -0800)]
Add ReductionLayer to reduce any number of "tail" axes to a scalar value

Currently implements operations SUM, MEAN, ASUM (sum of absolute
values), and SUMSQ (sum of squares)

9 years agoMerge pull request #2082 from jeffdonahue/flatten-layer-axis
Jeff Donahue [Wed, 3 Jun 2015 03:01:48 +0000 (20:01 -0700)]
Merge pull request #2082 from jeffdonahue/flatten-layer-axis

FlattenLayer gets a FlattenParameter with an axis, end_axis

9 years agoFlattenLayer gets a FlattenParameter with an axis, end_axis
Jeff Donahue [Thu, 1 Jan 2015 02:02:12 +0000 (18:02 -0800)]
FlattenLayer gets a FlattenParameter with an axis, end_axis

9 years agoMerge pull request #2090 from jeffdonahue/log-layer
Jeff Donahue [Wed, 3 Jun 2015 02:17:21 +0000 (19:17 -0700)]
Merge pull request #2090 from jeffdonahue/log-layer

LogLayer

9 years agoAdd LogLayer
Jeff Donahue [Fri, 2 Jan 2015 07:07:44 +0000 (23:07 -0800)]
Add LogLayer

9 years agoMerge pull request #2054 from mtamburrano/filter_layer_rebased
Jeff Donahue [Wed, 3 Jun 2015 01:32:01 +0000 (18:32 -0700)]
Merge pull request #2054 from mtamburrano/filter_layer_rebased

9 years agoFilterLayer cleanup and bugfix for GPU backward
Jeff Donahue [Wed, 3 Jun 2015 01:21:15 +0000 (18:21 -0700)]
FilterLayer cleanup and bugfix for GPU backward

-caffe_set -> caffe_gpu_set (backward was segfaulting before)
-remove uses of 'offset' (to support >4D blobs)
-change var++ -> ++var (per Google style guide)
-cleanup comments/whitespace

9 years agoFilter Layer implemented
manuele [Fri, 29 May 2015 13:54:38 +0000 (15:54 +0200)]
Filter Layer implemented

9 years agoMerge pull request #2532 from shelhamer/accum-prelu
Jeff Donahue [Tue, 2 Jun 2015 22:23:31 +0000 (15:23 -0700)]
Merge pull request #2532 from shelhamer/accum-prelu

PReLU accumulates grad

9 years agoMerge pull request #2536 from kibum14/master
Jeff Donahue [Tue, 2 Jun 2015 22:23:00 +0000 (15:23 -0700)]
Merge pull request #2536 from kibum14/master

Remove unnecessary filler parameter in the sample model

9 years agoRemove unnecessary filler parameter in the sample model
Kibum Bae [Tue, 2 Jun 2015 13:20:27 +0000 (22:20 +0900)]
Remove unnecessary filler parameter in the sample model

In bvlc_googlenet model, unnecessary filler parameters, 'std', were used
for Xavier weight fillers.

9 years agoPReLU accumulates grad
Takuya Narihira [Fri, 27 Feb 2015 02:59:27 +0000 (18:59 -0800)]
PReLU accumulates grad

9 years agoMerge pull request #2245 from dwt/add_homebrew_comments_to_makefile_config_example
Evan Shelhamer [Mon, 1 Jun 2015 02:33:44 +0000 (19:33 -0700)]
Merge pull request #2245 from dwt/add_homebrew_comments_to_makefile_config_example

[build] add commented out helpers for homebrew users

9 years agoMerge pull request #2064 from drohit92/patch-1
Evan Shelhamer [Mon, 1 Jun 2015 02:19:44 +0000 (19:19 -0700)]
Merge pull request #2064 from drohit92/patch-1

[docs] fix solverstate filename in ImageNet example

9 years agoMerge pull request #2163 from jeffdonahue/slice-concat-kernels
Evan Shelhamer [Mon, 1 Jun 2015 02:17:53 +0000 (19:17 -0700)]
Merge pull request #2163 from jeffdonahue/slice-concat-kernels

CUDA kernels for {Slice,Concat}Layer

9 years agoMerge pull request #2107 from 5kg/fix_mac_doc
Evan Shelhamer [Mon, 1 Jun 2015 02:16:29 +0000 (19:16 -0700)]
Merge pull request #2107 from 5kg/fix_mac_doc

[docs] fix OS X homebrew commands

9 years agoMerge pull request #2400 from cnDelbert/master
Evan Shelhamer [Mon, 1 Jun 2015 02:11:52 +0000 (19:11 -0700)]
Merge pull request #2400 from cnDelbert/master

Correct the REPO_DIRNAME in `examples/web_demo/app.py`

9 years agoMerge pull request #2350 from drdan14/log-parser-python-improved
Evan Shelhamer [Sat, 30 May 2015 07:26:27 +0000 (00:26 -0700)]
Merge pull request #2350 from drdan14/log-parser-python-improved

Python log parser improvements

9 years agoMerge pull request #2522 from MartinThoma/moose
Evan Shelhamer [Sat, 30 May 2015 07:25:07 +0000 (00:25 -0700)]
Merge pull request #2522 from MartinThoma/moose

[pycaffe] groom caffe.draw and draw_net.py

9 years agoMerge pull request #2498 from flx42/cpp_classification_example
Evan Shelhamer [Sat, 30 May 2015 07:22:35 +0000 (00:22 -0700)]
Merge pull request #2498 from flx42/cpp_classification_example

Add a simple C++ classification example

9 years agoMerge pull request #2528 from shelhamer/travis-lmdb-git
Evan Shelhamer [Sat, 30 May 2015 06:58:59 +0000 (23:58 -0700)]
Merge pull request #2528 from shelhamer/travis-lmdb-git

Fix lmdb install for travis

9 years agoMerge pull request #2527 from ronghanghu/matcaffe-update-mean-format
Evan Shelhamer [Sat, 30 May 2015 06:52:15 +0000 (23:52 -0700)]
Merge pull request #2527 from ronghanghu/matcaffe-update-mean-format

switch ilsvrc_2012_mean.mat to W x H x C, update demo and add comments

9 years ago[travis] install lmdb through git mirror
Evan Shelhamer [Sat, 30 May 2015 06:33:26 +0000 (23:33 -0700)]
[travis] install lmdb through git mirror

switch to github mirror of latest tag as tarball to fix stale url.

9 years agoUpdate ilsvrc_2012_mean.mat to W x H x C, update demo and add comments
Ronghang Hu [Sat, 30 May 2015 01:37:32 +0000 (09:37 +0800)]
Update ilsvrc_2012_mean.mat to W x H x C, update demo and add comments

Update previously ilsvrc_2012_mean.mat stores 'image_mean' variable in
H x W x C with BGR channels, which is inconsistent with Caffe's data format
and inconsistent with caffe.io.read_mean(..). Replace 'image_mean' with
'mean_data' variable in W x H x C and update classification_demo.m. Also add
some comments.

9 years agoMerge pull request #2511 from flx42/fix_illegal_mode_changes
Evan Shelhamer [Sat, 30 May 2015 06:11:35 +0000 (23:11 -0700)]
Merge pull request #2511 from flx42/fix_illegal_mode_changes

Fix invalid mode changes during tests

9 years agoMerge pull request #1977 from shelhamer/accum-grad
Evan Shelhamer [Sat, 30 May 2015 05:50:16 +0000 (22:50 -0700)]
Merge pull request #1977 from shelhamer/accum-grad

Decouple the computational batch size and minibatch size by accumulating gradients

9 years agoAdd a simple C++ classification example
Felix Abecassis [Sat, 30 May 2015 05:32:25 +0000 (22:32 -0700)]
Add a simple C++ classification example

Closes #2487

Example usage:
./build/examples/cpp_classification/classification.bin \
  models/bvlc_reference_caffenet/deploy.prototxt \
  models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel \
  data/ilsvrc12/imagenet_mean.binaryproto \
  data/ilsvrc12/synset_words.txt \
  examples/images/cat.jpg

9 years agoMerge pull request #2410 from sguada/datum_transform
Jeff Donahue [Sat, 30 May 2015 04:35:30 +0000 (21:35 -0700)]
Merge pull request #2410 from sguada/datum_transform

Datum transform

9 years ago Merge pull request #2294 from TorosFanny/master
Evan Shelhamer [Sat, 30 May 2015 04:23:22 +0000 (21:23 -0700)]
 Merge pull request #2294 from TorosFanny/master

  [example] fix path for diff in net surgery

9 years ago[example] fix path for diff in net surgery
TorosFanny [Sat, 30 May 2015 04:19:56 +0000 (21:19 -0700)]
[example] fix path for diff in net surgery

9 years agoMerge pull request #2240 from nsubtil/cmake-build-dependencies
Evan Shelhamer [Sat, 30 May 2015 01:23:42 +0000 (18:23 -0700)]
Merge pull request #2240 from nsubtil/cmake-build-dependencies

Wrangle (some) Caffe dependencies through CMake

9 years agoMerge pull request #2468 from Nerei/feature/minor_fix_in_cmake_config_generation
Evan Shelhamer [Sat, 30 May 2015 01:10:18 +0000 (18:10 -0700)]
Merge pull request #2468 from Nerei/feature/minor_fix_in_cmake_config_generation

[build] minor Cmake fix to clear python / numpy in CaffeConfig.cmake generation

9 years agoMerge pull request #2493 from longjon/sketchy-cuda-kernel-loop
Evan Shelhamer [Sat, 30 May 2015 01:08:27 +0000 (18:08 -0700)]
Merge pull request #2493 from longjon/sketchy-cuda-kernel-loop

Fix dangerous state in pooling and LRN CUDA kernels -- thanks @gustavla for the report in #2145

9 years agoMerge pull request #2514 from norouzi/master
Evan Shelhamer [Sat, 30 May 2015 00:56:09 +0000 (17:56 -0700)]
Merge pull request #2514 from norouzi/master

[bug] fix extract_features: zero pad keys + fix multi-feature dbtype bug

9 years agoMerge pull request #2505 from ronghanghu/matcaffe3
Evan Shelhamer [Fri, 29 May 2015 08:35:35 +0000 (01:35 -0700)]
Merge pull request #2505 from ronghanghu/matcaffe3

MatCaffe: overhaul and improve the MATLAB interface

9 years agoMore tests for Blob, Layer, copy_from and step, fix some typos
Ronghang Hu [Thu, 28 May 2015 23:50:23 +0000 (07:50 +0800)]
More tests for Blob, Layer, copy_from and step, fix some typos

More testes are added into test_net.m and test_solver.m

9 years agoFix automatic header file dependency for MatCaffe
Ronghang Hu [Thu, 28 May 2015 16:23:06 +0000 (00:23 +0800)]
Fix automatic header file dependency for MatCaffe

Automatic header file dependency was introduced in #1472, but
not correctly applied to matcaffe. Fix it by moving ./caffe_.d
to build/matlab/+caffe/private/caffe_.d and add it to DEPS

9 years agoMove demo to demo/ and check weights file existence
Ronghang Hu [Thu, 28 May 2015 10:33:54 +0000 (18:33 +0800)]
Move demo to demo/ and check weights file existence

Move all Matlab demo to caffe/matlab/demo. Since we want the user to add
caffe/matlab to Matlab search PATH, we don't want to mess it up with too
many files

Check if CaffeNet is already downloaded in classification demo.

9 years agoClean up old matcaffe wrapper and rename caffe.reset to caffe.reset_all
Ronghang Hu [Thu, 28 May 2015 08:24:30 +0000 (16:24 +0800)]
Clean up old matcaffe wrapper and rename caffe.reset to caffe.reset_all

Remove old matlab wrapper but keep the classification demo and hdf5 demo

Change 'caffe.reset()' to 'caffe.reset_all()' to avoid potential name conflict.
Otherwise, Matlab R2015a complains:
Warning: Function reset has the same name as a MATLAB builtin. We suggest you rename the
function to avoid a potential name conflict.

9 years agoAdd MatCaffe docs to docs/tutorial/interfaces.md
Ronghang Hu [Thu, 28 May 2015 06:52:47 +0000 (14:52 +0800)]
Add MatCaffe docs to docs/tutorial/interfaces.md

9 years agoAesthetic changes on code style and some minor fix
Ronghang Hu [Thu, 28 May 2015 05:40:26 +0000 (13:40 +0800)]
Aesthetic changes on code style and some minor fix

9 years agoFix matlab tailing dimension 1 issue for shape match
Ronghang Hu [Wed, 27 May 2015 04:37:20 +0000 (12:37 +0800)]
Fix matlab tailing dimension 1 issue for shape match

Matlab cannot have tailing dimension 1 for ndim > 2, so you
cannot create 20 x 10 x 1 x 1 array in matlab as it becomes 20 x 10.
Extend matlab arrays to have tailing dimension 1 during shape match.

9 years agoMatCaffe3 : a powerful matlab interface for caffe
Ronghang Hu [Thu, 21 May 2015 17:45:14 +0000 (01:45 +0800)]
MatCaffe3 : a powerful matlab interface for caffe

Added matcaffe3, a powerful matlab interface. To test it, run 'make mattest'

9 years agodirectly normalize accumulated gradients
Evan Shelhamer [Thu, 28 May 2015 19:43:29 +0000 (12:43 -0700)]
directly normalize accumulated gradients

`SGDSolver::Normalize()` normalizes accumulated gradients by scaling
inversely to the accumulation as `1 / iter_size`.

This fixes accumulation for AdaGrad and is more obvious than fooling
with rates and decays in 55585f5.

9 years agotest equivalence of solving with accumulating gradients
Evan Shelhamer [Fri, 22 May 2015 01:14:16 +0000 (18:14 -0700)]
test equivalence of solving with accumulating gradients

Compare the parameters after solving with a given batch size and the
halved batch size + two iter accumulation of gradients equivalent.

Note: the test net dummy data layer now makes constant data and random
gaussian targets. This assures the standard and gradient accumulation
cases check the same data. Otherwise the difference in batch sizes
causes different orders of random number draws.

9 years agopython/draw_net.py and python/caffe/draw.py: Simplified code; added more docstrings...
Martin Thoma [Wed, 27 May 2015 21:20:16 +0000 (23:20 +0200)]
python/draw_net.py and python/caffe/draw.py: Simplified code; added more docstrings; adjusted code according to PEP8

9 years agoadjust local learning rate and decay according to gradient accumulation
Evan Shelhamer [Fri, 22 May 2015 00:06:42 +0000 (17:06 -0700)]
adjust local learning rate and decay according to gradient accumulation

Divide local rate by `iter_size` to normalize the gradient according to
the full minibatch size and not only the computational batch size.

Multiply the local decay by `iter_size` to counter the division of the
local learning rate since the decay is multiplied by the rate in the
update equation.

9 years agoaccumulate gradients in cudnn conv layer
Jonathan L Long [Sun, 14 Sep 2014 00:41:59 +0000 (17:41 -0700)]
accumulate gradients in cudnn conv layer

9 years agoaccumulate gradients in (de)conv layers
Jonathan L Long [Wed, 31 Dec 2014 06:29:35 +0000 (22:29 -0800)]
accumulate gradients in (de)conv layers

9 years agoaccumulate gradients in inner product layer
Sergio [Sat, 27 Sep 2014 06:03:26 +0000 (23:03 -0700)]
accumulate gradients in inner product layer

9 years agozero-init param diffs in gradient checker
Jonathan L Long [Wed, 31 Dec 2014 06:52:07 +0000 (22:52 -0800)]
zero-init param diffs in gradient checker

9 years agozero-init param diffs and accumulate gradients
Jonathan L Long [Tue, 12 Aug 2014 04:38:59 +0000 (21:38 -0700)]
zero-init param diffs and accumulate gradients

(With layers whose backward accumulates gradients), this effectively
decouples the computational batch from the SGD minibatch. Each
iteration accumulates gradients over iter_size batches, then parameters
are updated.

9 years agoMerge pull request #2518 from shelhamer/dedup_solvers
Jeff Donahue [Wed, 27 May 2015 20:42:47 +0000 (13:42 -0700)]
Merge pull request #2518 from shelhamer/dedup_solvers

Deduplicate solver regularization, logging, and local rates and decays

9 years agoSolver::MakeUpdate() -> Solver::ApplyUpdate
Evan Shelhamer [Wed, 27 May 2015 19:24:06 +0000 (12:24 -0700)]
Solver::MakeUpdate() -> Solver::ApplyUpdate

Designate `Solver::ApplyUpdate()` as the core method to compute
and apply parameter updates given the current state of the Net.

Make `Solver::ComputeUpdateValue()` a subordinate call overloaded by the
`SGDSolver`s to take care of optimization algorithm details.

9 years agofix the bug with db_type when the number of features to be extracted is larger than 1
Mohammad Norouzi [Wed, 27 May 2015 14:33:19 +0000 (10:33 -0400)]
fix the bug with db_type when the number of features to be extracted is larger than 1

9 years agodeduplicate decay and local rate in solver updates
Evan Shelhamer [Thu, 21 May 2015 23:34:43 +0000 (16:34 -0700)]
deduplicate decay and local rate in solver updates

9 years agoRefactor solvers regularization and logging code
Cyprien Noel [Tue, 19 May 2015 01:30:00 +0000 (18:30 -0700)]
Refactor solvers regularization and logging code

9 years agoadd leading zeros to keys in feature DB files
Mohammad Norouzi [Tue, 26 May 2015 21:25:56 +0000 (17:25 -0400)]
add leading zeros to keys in feature DB files

9 years agoMake class MultinomialLogisticLossLayerTest derive from CPUDeviceTest
Felix Abecassis [Tue, 26 May 2015 18:21:58 +0000 (11:21 -0700)]
Make class MultinomialLogisticLossLayerTest derive from CPUDeviceTest

9 years agoMake class CuDNNSoftmaxLayerTest derive from GPUDeviceTest
Felix Abecassis [Tue, 26 May 2015 18:21:56 +0000 (11:21 -0700)]
Make class CuDNNSoftmaxLayerTest derive from GPUDeviceTest

9 years agoMake class DummyDataLayerTest derive from CPUDeviceTest
Felix Abecassis [Tue, 26 May 2015 18:21:53 +0000 (11:21 -0700)]
Make class DummyDataLayerTest derive from CPUDeviceTest

9 years agoMake class CuDNNPoolingLayerTest derive from GPUDeviceTest
Felix Abecassis [Tue, 26 May 2015 18:21:50 +0000 (11:21 -0700)]
Make class CuDNNPoolingLayerTest derive from GPUDeviceTest

9 years agoMake class CuDNNConvolutionLayerTest derive from GPUDeviceTest
Felix Abecassis [Tue, 26 May 2015 18:21:48 +0000 (11:21 -0700)]
Make class CuDNNConvolutionLayerTest derive from GPUDeviceTest

9 years agoMake class ArgMaxLayerTest derive from CPUDeviceTest
Felix Abecassis [Tue, 26 May 2015 18:21:46 +0000 (11:21 -0700)]
Make class ArgMaxLayerTest derive from CPUDeviceTest

9 years agoMake class CuDNNNeuronLayerTest derive from GPUDeviceTest
Felix Abecassis [Tue, 26 May 2015 18:21:44 +0000 (11:21 -0700)]
Make class CuDNNNeuronLayerTest derive from GPUDeviceTest

9 years agoMake class AccuracyLayerTest derive from CPUDeviceTest
Felix Abecassis [Tue, 26 May 2015 18:21:41 +0000 (11:21 -0700)]
Make class AccuracyLayerTest derive from CPUDeviceTest

9 years agoMake class Im2colKernelTest derive from GPUDeviceTest
Felix Abecassis [Tue, 26 May 2015 18:21:39 +0000 (11:21 -0700)]
Make class Im2colKernelTest derive from GPUDeviceTest

9 years agoAdd classes GPUDeviceTest and CPUDeviceTest.
Felix Abecassis [Tue, 26 May 2015 18:21:36 +0000 (11:21 -0700)]
Add classes GPUDeviceTest and CPUDeviceTest.

These new classes can be used to implement test cases that are only
running on the GPU or the CPU. The goal is to move all calls to
Caffe::set_mode() inside the test framework, to discourage any test to
change the mode halfway through the execution, which is documented to
be illegal.

9 years agoSplit class StochasticPoolingLayerTest into CPUStochasticPoolingLayerTest and GPUStoc...
Felix Abecassis [Tue, 26 May 2015 18:21:32 +0000 (11:21 -0700)]
Split class StochasticPoolingLayerTest into CPUStochasticPoolingLayerTest and GPUStochasticPoolingLayerTest

9 years agoMerge pull request #1946 from nickcarlevaris/msra_init
Evan Shelhamer [Tue, 26 May 2015 20:13:01 +0000 (13:13 -0700)]
Merge pull request #1946 from nickcarlevaris/msra_init

  Add MSRAFiller, an Xavier-like filler designed for use with ReLUs

9 years agoinclude comment on Saxe and sqrt(2) scaling factor
Evan Shelhamer [Tue, 26 May 2015 19:39:14 +0000 (12:39 -0700)]
include comment on Saxe and sqrt(2) scaling factor

although different and independent, the derivation of Saxe et
al. with regards to the scaling factor might be of interest.

9 years agoAdded MSRAFiller, an Xavier-like filler designed for use with ReLUs
Nick Carlevaris-Bianco [Mon, 16 Feb 2015 05:19:43 +0000 (15:49 +1030)]
Added MSRAFiller, an Xavier-like filler designed for use with ReLUs

...instead of tanh. Based on paper: He et al, "Delving Deep into
Rectifiers: Surpassing Human-Level Performance on ImageNet
Classification," 2015.

- add VarianceNorm option to FillerParameters which allows one to
  normalize by fan_in, fan_out or their average.
- update XavierFiller to use the VarianceNorm option (default behavior
  unchanged).
- add tests for MSRAFiller and XavierFiller.

9 years agoSplit class MathFunctionsTest into CPUMathFunctionsTest and GPUMathFunctionsTest
Felix Abecassis [Tue, 26 May 2015 18:21:28 +0000 (11:21 -0700)]
Split class MathFunctionsTest into CPUMathFunctionsTest and GPUMathFunctionsTest

9 years agoRefactor types FloatCPU and DoubleCPU into a new type CPUDevice<T>
Felix Abecassis [Tue, 26 May 2015 18:21:15 +0000 (11:21 -0700)]
Refactor types FloatCPU and DoubleCPU into a new type CPUDevice<T>

Similarly, FloatGPU and DoubleGPU are replaced by a new type GPUDevice<T>.