platform/upstream/caffeonacl.git
10 years agoset seed in neuron and power layer tests for deterministic results
Jeff Donahue [Sat, 26 Apr 2014 22:47:35 +0000 (15:47 -0700)]
set seed in neuron and power layer tests for deterministic results

10 years agoMerge pull request #367 from shelhamer/randomize-test-order
Jeff Donahue [Sat, 26 Apr 2014 22:45:11 +0000 (15:45 -0700)]
Merge pull request #367 from shelhamer/randomize-test-order

Randomize order of test execution by `make runtest`

10 years agoMerge pull request #303 from longjon/hinge-loss-layer
Jeff Donahue [Sat, 26 Apr 2014 22:13:15 +0000 (15:13 -0700)]
Merge pull request #303 from longjon/hinge-loss-layer

HingeLossLayer

10 years agoMerge pull request #368 from shelhamer/fix-data-layer-sequence-tests
Jeff Donahue [Sat, 26 Apr 2014 22:09:17 +0000 (15:09 -0700)]
Merge pull request #368 from shelhamer/fix-data-layer-sequence-tests

Fix data layer sequence tests to avoid leveldb contention and lock failures

10 years agoscope data layer sequence tests to avoid leveldb contention
Evan Shelhamer [Sat, 26 Apr 2014 21:57:11 +0000 (14:57 -0700)]
scope data layer sequence tests to avoid leveldb contention

10 years agoset phase in dropout tests
Evan Shelhamer [Sat, 26 Apr 2014 21:42:38 +0000 (14:42 -0700)]
set phase in dropout tests

10 years agorandomize order of test execution by make runtest
Evan Shelhamer [Sat, 26 Apr 2014 01:54:12 +0000 (18:54 -0700)]
randomize order of test execution by make runtest

10 years agoMerge pull request #366 from shelhamer/test-phase-fix
Jeff Donahue [Sat, 26 Apr 2014 01:42:37 +0000 (18:42 -0700)]
Merge pull request #366 from shelhamer/test-phase-fix

set TRAIN in CommonTest.TestPhase

10 years agoset TRAIN in CommonTest.TestPhase
Evan Shelhamer [Sat, 26 Apr 2014 01:33:40 +0000 (18:33 -0700)]
set TRAIN in CommonTest.TestPhase

...to avoid effects of test execution order. The original intention was
to check that TRAIN is the default, but it is better to have consistent
results.

10 years agotest HingeLossLayer
Jonathan L Long [Fri, 25 Apr 2014 23:44:05 +0000 (16:44 -0700)]
test HingeLossLayer

Based on SoftmaxWithLossLayerTest.

10 years agomake gradient checker's kink use feature absolute value
Jonathan L Long [Fri, 25 Apr 2014 23:38:48 +0000 (16:38 -0700)]
make gradient checker's kink use feature absolute value

In theory, layer functions could be nonsmooth anywhere; in all cases in
use so far, they are nonsmooth at either zero or +1 and -1. In the
future, it might be necessary to generalize the kink mechanism beyond
this stopgap measure.

10 years agoadd HingeLossLayer for one-vs-all hinge loss
Jonathan L Long [Tue, 8 Apr 2014 03:35:33 +0000 (20:35 -0700)]
add HingeLossLayer for one-vs-all hinge loss

This layer implements a "one-vs-all" hinge loss, (1/n) sum_ij max(0, 1 -
y_ij x_ij), with bottom blob x_ij (i ranging over examples and j over
classes), and y_ij = +1/-1 indicating the label. No regularization is
included, since regularization is done via weight decay or using the
parameters of another layer. The gradient is taken to be zero at the
hinge point. This commit only provides the CPU implementation.

10 years agoMerge pull request #365 from longjon/pycaffe-empty-nets
Evan Shelhamer [Fri, 25 Apr 2014 23:02:37 +0000 (16:02 -0700)]
Merge pull request #365 from longjon/pycaffe-empty-nets

Add unary CaffeNet constructor for uninitialized nets

10 years agopycaffe: add unary CaffeNet constructor for uninitialized nets
Jonathan L Long [Fri, 25 Apr 2014 21:50:36 +0000 (14:50 -0700)]
pycaffe: add unary CaffeNet constructor for uninitialized nets

Note that parameters are uninitialized, not zero-filled.

10 years agoMerge pull request #363 from jeffdonahue/speedup-gradient-check
Evan Shelhamer [Fri, 25 Apr 2014 19:51:22 +0000 (12:51 -0700)]
Merge pull request #363 from jeffdonahue/speedup-gradient-check

Gradient check speedups

10 years agoeltwise gradient checker
Jeff Donahue [Fri, 25 Apr 2014 03:24:19 +0000 (20:24 -0700)]
eltwise gradient checker

10 years agomove analytic gradient computation outside loop and store -- saves a lot
Jeff Donahue [Fri, 25 Apr 2014 04:44:48 +0000 (21:44 -0700)]
move analytic gradient computation outside loop and store -- saves a lot
of time

10 years agoMerge pull request #364 from niuzhiheng/dev
Evan Shelhamer [Fri, 25 Apr 2014 16:23:22 +0000 (09:23 -0700)]
Merge pull request #364 from niuzhiheng/dev

Update drawnet.py to support new-style net definitions

10 years agonote pydot dependency of model visualization
Evan Shelhamer [Fri, 25 Apr 2014 16:23:00 +0000 (09:23 -0700)]
note pydot dependency of model visualization

10 years agoUpdate the drawnet.py to reflect the recent revised net definition.
ZhiHeng NIU [Fri, 25 Apr 2014 09:43:56 +0000 (17:43 +0800)]
Update the drawnet.py to reflect the recent revised net definition.

10 years agoMerge pull request #352 from jeffdonahue/solver-seed
Evan Shelhamer [Thu, 24 Apr 2014 16:37:43 +0000 (09:37 -0700)]
Merge pull request #352 from jeffdonahue/solver-seed

Eliminate use of "rand" in code; enable deterministic results via solver seed

10 years agoMerge pull request #357 from jeffdonahue/test-on-iter-0
Evan Shelhamer [Thu, 24 Apr 2014 16:35:13 +0000 (09:35 -0700)]
Merge pull request #357 from jeffdonahue/test-on-iter-0

Test before doing any training

10 years agotest on "0th iteration" -- before doing any training
Jeff Donahue [Thu, 24 Apr 2014 09:05:32 +0000 (02:05 -0700)]
test on "0th iteration" -- before doing any training

10 years agoadd caffe/random_fn lint rule to check for use of rand, rand_r, random
Jeff Donahue [Wed, 23 Apr 2014 01:00:22 +0000 (18:00 -0700)]
add caffe/random_fn lint rule to check for use of rand, rand_r, random

10 years agoreplace std::shuffle with version using prefetch rng; improve unit test
Jeff Donahue [Tue, 22 Apr 2014 23:32:40 +0000 (16:32 -0700)]
replace std::shuffle with version using prefetch rng; improve unit test

10 years agoreplace remaining uses of rand() with caffe_rng_rand()
Jeff Donahue [Tue, 22 Apr 2014 21:12:36 +0000 (14:12 -0700)]
replace remaining uses of rand() with caffe_rng_rand()

10 years agoprefetch_rng in window_data_layer
Jeff Donahue [Tue, 22 Apr 2014 21:09:37 +0000 (14:09 -0700)]
prefetch_rng in window_data_layer

10 years agoprefetch_rng in ImageDataLayer
Jeff Donahue [Tue, 22 Apr 2014 20:56:02 +0000 (13:56 -0700)]
prefetch_rng in ImageDataLayer

10 years agotest scale param
Jeff Donahue [Tue, 22 Apr 2014 20:17:05 +0000 (13:17 -0700)]
test scale param

10 years agomake seed test pass by setting up new layer to generate the 2nd sequence
Jeff Donahue [Tue, 22 Apr 2014 20:08:26 +0000 (13:08 -0700)]
make seed test pass by setting up new layer to generate the 2nd sequence

10 years agocleanup data_layer, add prefetch_rng_ field to it and use instead of
Jeff Donahue [Tue, 22 Apr 2014 19:41:02 +0000 (12:41 -0700)]
cleanup data_layer, add prefetch_rng_ field to it and use instead of
rand -- seeded tests still fail

10 years agoadd tests for random crop sequence -- currently both fail
Jeff Donahue [Tue, 22 Apr 2014 18:34:40 +0000 (11:34 -0700)]
add tests for random crop sequence -- currently both fail

10 years agoadd data layer crop tests
Jeff Donahue [Tue, 22 Apr 2014 11:42:47 +0000 (04:42 -0700)]
add data layer crop tests

10 years agoadd random_seed field to SolverParameter and have solver use it --
Jeff Donahue [Thu, 17 Apr 2014 04:00:30 +0000 (21:00 -0700)]
add random_seed field to SolverParameter and have solver use it --
already works for lenet, doesn't work for imagenet w/ rand() calls

10 years agoadd forward tests (via reference impl) for SigmoidCrossEntropyLayer
Jeff Donahue [Wed, 23 Apr 2014 06:12:14 +0000 (23:12 -0700)]
add forward tests (via reference impl) for SigmoidCrossEntropyLayer

10 years agoMerge pull request #350 from longjon/finicky-exhaustive
Evan Shelhamer [Tue, 22 Apr 2014 22:49:20 +0000 (15:49 -0700)]
Merge pull request #350 from longjon/finicky-exhaustive

Make CheckGradientExhaustive fail for topless layers

10 years agomake CheckGradientExhaustive fail for topless layers
Jonathan L Long [Tue, 22 Apr 2014 06:41:50 +0000 (23:41 -0700)]
make CheckGradientExhaustive fail for topless layers

Without this commit, it is possible to mistakenly call
CheckGradientExhaustive on a layer with no top blobs (i.e., a loss
layer), causing the gradient check to silently succeed while doing
nothing. With this commit, doing so will cause the test to fail.

10 years agoMerge pull request #348 from jeffdonahue/datalayer-singleton-bugfix
Evan Shelhamer [Tue, 22 Apr 2014 05:20:36 +0000 (22:20 -0700)]
Merge pull request #348 from jeffdonahue/datalayer-singleton-bugfix

Fix singleton call in data layers

Note previous crash reports were not in fact due to RNG, although they led to improvements of the RNG code.

10 years agodo the same as prev commit for ImageDataLayer
Jeff Donahue [Tue, 22 Apr 2014 03:43:39 +0000 (20:43 -0700)]
do the same as prev commit for ImageDataLayer

10 years agofix bug where DataLayerPrefetch creates its own Caffe singleton, causing
Jeff Donahue [Tue, 22 Apr 2014 03:35:43 +0000 (20:35 -0700)]
fix bug where DataLayerPrefetch creates its own Caffe singleton, causing
the phase to always be set to TRAIN (always random crops) and RNG failures

10 years agoComment current forward/backward responsibilities
Evan Shelhamer [Mon, 21 Apr 2014 03:12:57 +0000 (20:12 -0700)]
Comment current forward/backward responsibilities

10 years agoProofread install docs
Evan Shelhamer [Mon, 21 Apr 2014 03:07:06 +0000 (20:07 -0700)]
Proofread install docs

10 years agoMerge pull request #339 from sergeyk/dev
Evan Shelhamer [Sun, 20 Apr 2014 23:57:56 +0000 (16:57 -0700)]
Merge pull request #339 from sergeyk/dev

Update installation docs

10 years agoinstallation doc update
Sergey Karayev [Sun, 20 Apr 2014 20:38:30 +0000 (13:38 -0700)]
installation doc update

10 years agoMerge pull request #336 from jeffdonahue/fix-rng-segfault
Evan Shelhamer [Sat, 19 Apr 2014 06:18:59 +0000 (23:18 -0700)]
Merge pull request #336 from jeffdonahue/fix-rng-segfault

Fix RNG segfault related to #297

10 years agofix examples path in mnist leveldb sh
Evan Shelhamer [Fri, 18 Apr 2014 18:51:41 +0000 (11:51 -0700)]
fix examples path in mnist leveldb sh

10 years agoMerge pull request #332 from jeffdonahue/share-trained-layers
Jeff Donahue [Fri, 18 Apr 2014 18:11:01 +0000 (11:11 -0700)]
Merge pull request #332 from jeffdonahue/share-trained-layers

add ShareTrainedLayersWith method and use for test net in solver

10 years agoremove now unused set_generator and related code
Jeff Donahue [Fri, 18 Apr 2014 17:51:30 +0000 (10:51 -0700)]
remove now unused set_generator and related code

10 years agopass caffe rng ref into variate_generator constructor instead of having
Jeff Donahue [Fri, 18 Apr 2014 17:36:28 +0000 (10:36 -0700)]
pass caffe rng ref into variate_generator constructor instead of having
caffe rng adopt its state

10 years agoremove unnecessary return from void set_generator
Jeff Donahue [Fri, 18 Apr 2014 01:23:33 +0000 (18:23 -0700)]
remove unnecessary return from void set_generator

10 years agoadd ShareTrainedLayersWith method and use for test net in solver
Jeff Donahue [Wed, 16 Apr 2014 17:01:30 +0000 (10:01 -0700)]
add ShareTrainedLayersWith method and use for test net in solver

10 years agoMerge pull request #330 from jeffdonahue/mnist-autoencoder-example
Evan Shelhamer [Wed, 16 Apr 2014 15:46:49 +0000 (09:46 -0600)]
Merge pull request #330 from jeffdonahue/mnist-autoencoder-example

MNIST autoencoder example

10 years agochange to correct next layer id for merge
Jeff Donahue [Wed, 16 Apr 2014 15:14:27 +0000 (08:14 -0700)]
change to correct next layer id for merge

10 years agoclear sigmoid top vec at initialization
Jeff Donahue [Wed, 16 Apr 2014 15:13:20 +0000 (08:13 -0700)]
clear sigmoid top vec at initialization

10 years agoadd sigmoid cross ent layer unit tests
Jeff Donahue [Wed, 16 Apr 2014 05:46:45 +0000 (22:46 -0700)]
add sigmoid cross ent layer unit tests

10 years agomnist_autoencoder_solver cleanup
Jeff Donahue [Wed, 16 Apr 2014 05:26:11 +0000 (22:26 -0700)]
mnist_autoencoder_solver cleanup

10 years agochange lenet dir to 'mnist' in docs
Jeff Donahue [Wed, 16 Apr 2014 05:25:46 +0000 (22:25 -0700)]
change lenet dir to 'mnist' in docs

10 years agomake solver able to compute and display test loss
Jeff Donahue [Tue, 15 Apr 2014 22:38:57 +0000 (15:38 -0700)]
make solver able to compute and display test loss

10 years agomnist autoencoder test proto bugfix: add sigmoid layer before loss
Jeff Donahue [Tue, 15 Apr 2014 22:38:06 +0000 (15:38 -0700)]
mnist autoencoder test proto bugfix: add sigmoid layer before loss

10 years agoenable DataLayer to output unlabeled data
Jeff Donahue [Tue, 15 Apr 2014 22:12:57 +0000 (15:12 -0700)]
enable DataLayer to output unlabeled data

10 years agoadd mnist autoencoder example necessities (sigmoid cross entropy loss
Jeff Donahue [Tue, 15 Apr 2014 21:52:41 +0000 (14:52 -0700)]
add mnist autoencoder example necessities (sigmoid cross entropy loss
layer, sparse gaussian filler)

10 years agorename lenet dir to mnist
Jeff Donahue [Tue, 15 Apr 2014 17:56:32 +0000 (10:56 -0700)]
rename lenet dir to mnist

10 years agoGive choice of ATLAS, MKL, and OpenBLAS (with option to override paths)
Evan Shelhamer [Tue, 15 Apr 2014 07:37:06 +0000 (01:37 -0600)]
Give choice of ATLAS, MKL, and OpenBLAS (with option to override paths)

- configure build for ATLAS, MKL, or OpenBLAS on Linux and OSX
- allow overriding of the include or lib dirs
- replace magic numbers with BLAS names (atlas, mkl, open)

Follow-up from #305 and #325.

10 years agorename python include config var to match lib
Evan Shelhamer [Tue, 15 Apr 2014 07:36:18 +0000 (01:36 -0600)]
rename python include config var to match lib

10 years agoMerge pull request #325 from AlOa/OpenBlas
Evan Shelhamer [Tue, 15 Apr 2014 07:47:57 +0000 (01:47 -0600)]
Merge pull request #325 from AlOa/OpenBlas

Add possibility to use OpenBlas

10 years agoAdd possibility to use OpenBlas
AlOa [Mon, 14 Apr 2014 14:12:12 +0000 (16:12 +0200)]
Add possibility to use OpenBlas

10 years agoMerge pull request #318 from jeffdonahue/blob-copy-by-reference
Evan Shelhamer [Sun, 13 Apr 2014 01:40:08 +0000 (18:40 -0700)]
Merge pull request #318 from jeffdonahue/blob-copy-by-reference

add Share{Data,Diff} methods to blobs to enable "virtual" copies

- reduce operations by not copying values
- spare memory by shared_ptrs
- simplify SplitLayer logic

10 years agoMerge pull request #319 from jeffdonahue/sigmoid-optimization
Evan Shelhamer [Sun, 13 Apr 2014 01:26:30 +0000 (18:26 -0700)]
Merge pull request #319 from jeffdonahue/sigmoid-optimization

sigmoid layer backward pass optimization: don't recompute forward pass

10 years agosigmoid layer backward pass optimization: don't recompute forward pass
Jeff Donahue [Sat, 12 Apr 2014 11:16:23 +0000 (04:16 -0700)]
sigmoid layer backward pass optimization: don't recompute forward pass

10 years agochange Adopt -> Share as suggested by kloudkl
Jeff Donahue [Sat, 12 Apr 2014 07:52:15 +0000 (00:52 -0700)]
change Adopt -> Share as suggested by kloudkl

10 years agoadd Adopt{Data,Diff} methods to blobs to enable "virtual copying"
Jeff Donahue [Sat, 12 Apr 2014 01:07:08 +0000 (18:07 -0700)]
add Adopt{Data,Diff} methods to blobs to enable "virtual copying"

10 years agoadd unit tests for cpu/gpu copy functions
Jeff Donahue [Fri, 11 Apr 2014 23:07:43 +0000 (16:07 -0700)]
add unit tests for cpu/gpu copy functions

10 years agochange some unnecessary TYPED_TESTs to TEST_Fs
Jeff Donahue [Sat, 12 Apr 2014 06:47:40 +0000 (23:47 -0700)]
change some unnecessary TYPED_TESTs to TEST_Fs

10 years agofix lint errors by adding 'explicit' to new single arg pycaffe
Jeff Donahue [Fri, 11 Apr 2014 23:10:15 +0000 (16:10 -0700)]
fix lint errors by adding 'explicit' to new single arg pycaffe
constructors

10 years agopolished ignore
Evan Shelhamer [Fri, 11 Apr 2014 05:49:50 +0000 (22:49 -0700)]
polished ignore

10 years agoBack-merge docs and example image changes from `master` to `dev`
Evan Shelhamer [Thu, 10 Apr 2014 00:11:49 +0000 (17:11 -0700)]
Back-merge docs and example image changes from `master` to `dev`

10 years agoMerge pull request #310 from jeffdonahue/clang-test-power-layer-fixes
Evan Shelhamer [Wed, 9 Apr 2014 22:54:40 +0000 (15:54 -0700)]
Merge pull request #310 from jeffdonahue/clang-test-power-layer-fixes

clang fixes for test_power_layer

10 years agoadd using std::isnan and use this-> when calling Test{For,Back}ward
Jeff Donahue [Wed, 9 Apr 2014 22:45:37 +0000 (15:45 -0700)]
add using std::isnan and use this-> when calling Test{For,Back}ward

10 years agofix osx 10.9 condition in Makefile
Evan Shelhamer [Wed, 9 Apr 2014 22:44:07 +0000 (15:44 -0700)]
fix osx 10.9 condition in Makefile

10 years agoinclude vecLib BLAS dir on osx
Evan Shelhamer [Wed, 9 Apr 2014 21:33:31 +0000 (14:33 -0700)]
include vecLib BLAS dir on osx

10 years agoMerge pull request #286 from longjon/pycaffe-solver
Evan Shelhamer [Wed, 9 Apr 2014 20:31:18 +0000 (13:31 -0700)]
Merge pull request #286 from longjon/pycaffe-solver

Expose SGDSolver to pycaffe, switch from inheritance to overriding for caffe.Net

10 years agochange true_std to intended bernoulli_std
Jeff Donahue [Wed, 9 Apr 2014 18:11:04 +0000 (11:11 -0700)]
change true_std to intended bernoulli_std

10 years agoMerge pull request #297 from jeffdonahue/rng-bug
Evan Shelhamer [Wed, 9 Apr 2014 18:04:12 +0000 (11:04 -0700)]
Merge pull request #297 from jeffdonahue/rng-bug

Fix persistence of random state with Boost RNG

10 years agomore rng test cleanup
Jeff Donahue [Wed, 9 Apr 2014 02:45:31 +0000 (19:45 -0700)]
more rng test cleanup

10 years agore-time imagenet example on k20, instead of my laptop
Evan Shelhamer [Wed, 9 Apr 2014 04:52:08 +0000 (21:52 -0700)]
re-time imagenet example on k20, instead of my laptop

10 years agochange *Plus* tests to *Times* tests because the Plus tests don't
Jeff Donahue [Wed, 9 Apr 2014 00:15:28 +0000 (17:15 -0700)]
change *Plus* tests to *Times* tests because the Plus tests don't
actually check for uncorrelated RNG results

10 years agoadd analogous caffe_gpu_rng_* functions for gaussian and uniform, and
Jeff Donahue [Tue, 8 Apr 2014 23:24:50 +0000 (16:24 -0700)]
add analogous caffe_gpu_rng_* functions for gaussian and uniform, and
add test cases

10 years agohave rng_stream initialize RNG if not already initialized
Jeff Donahue [Tue, 8 Apr 2014 21:24:48 +0000 (14:24 -0700)]
have rng_stream initialize RNG if not already initialized

10 years agomake RNG function outputs the last argument per Google C++ style guidelines
Jeff Donahue [Tue, 8 Apr 2014 20:18:29 +0000 (13:18 -0700)]
make RNG function outputs the last argument per Google C++ style guidelines

10 years agomake RNG function names more similar to other caffe math function names
Jeff Donahue [Tue, 8 Apr 2014 18:57:25 +0000 (11:57 -0700)]
make RNG function names more similar to other caffe math function names

10 years agogpu_hamming_distance fails unit test with fixed RNG; mark it NOT_IMPLEMENTED
Jeff Donahue [Tue, 8 Apr 2014 11:25:20 +0000 (04:25 -0700)]
gpu_hamming_distance fails unit test with fixed RNG; mark it NOT_IMPLEMENTED
with TODO to fix and disable its unit test

10 years agocleanup RNG unit tests
Jeff Donahue [Tue, 8 Apr 2014 10:54:27 +0000 (03:54 -0700)]
cleanup RNG unit tests

10 years agomake rng_ a private member of Generator
Jeff Donahue [Tue, 8 Apr 2014 04:16:00 +0000 (21:16 -0700)]
make rng_ a private member of Generator

10 years agocleanup test_math_functions
Jeff Donahue [Tue, 8 Apr 2014 02:05:59 +0000 (19:05 -0700)]
cleanup test_math_functions

10 years agocomment to explain the purpose of Caffe::set_generator
Jeff Donahue [Tue, 8 Apr 2014 02:08:27 +0000 (19:08 -0700)]
comment to explain the purpose of Caffe::set_generator

10 years agofix bernoulli*bernoulli test, now all pass
Jeff Donahue [Tue, 8 Apr 2014 00:58:11 +0000 (17:58 -0700)]
fix bernoulli*bernoulli test, now all pass

10 years agocall caffe_set_rng at the end of each vRng function to maintain state
Jeff Donahue [Tue, 8 Apr 2014 00:14:23 +0000 (17:14 -0700)]
call caffe_set_rng at the end of each vRng function to maintain state

10 years agoadd bernoulli*bernoulli test
Jeff Donahue [Mon, 7 Apr 2014 19:24:21 +0000 (12:24 -0700)]
add bernoulli*bernoulli test

10 years agocleanup log messages
Jeff Donahue [Mon, 7 Apr 2014 02:57:44 +0000 (19:57 -0700)]
cleanup log messages

10 years agoadd analogous test for uniform instead of gaussian
Jeff Donahue [Mon, 7 Apr 2014 02:41:14 +0000 (19:41 -0700)]
add analogous test for uniform instead of gaussian