Kai Li [Thu, 28 Aug 2014 19:02:10 +0000 (03:02 +0800)]
Add transformer to the memory data layer
Kai Li [Thu, 28 Aug 2014 09:45:39 +0000 (17:45 +0800)]
Implement Forward_gpu in the base prefetching data layer
Kai Li [Thu, 28 Aug 2014 09:32:46 +0000 (17:32 +0800)]
The BasePrefetchingDataLayer shouldn't join the thread
Kai Li [Thu, 28 Aug 2014 08:55:56 +0000 (16:55 +0800)]
Simplify the WindowDataLayer using the base class
Kai Li [Thu, 28 Aug 2014 08:22:41 +0000 (16:22 +0800)]
Remove duplicate codes from the ImageDataLayer
Kai Li [Thu, 28 Aug 2014 08:10:47 +0000 (16:10 +0800)]
Extract common data layer functionalities out of the DataLayer
Kai Li [Thu, 28 Aug 2014 07:29:25 +0000 (15:29 +0800)]
Create base data layer and base prefetching data layer
Evan Shelhamer [Tue, 2 Sep 2014 15:38:07 +0000 (08:38 -0700)]
Merge pull request #1025 from mohomran/minor_fix_to_mnist_solver_prototxt
Snapshot naming issue in mnist_autoencoder_solver_adagrad.prototxt
Mohamed Omran [Tue, 2 Sep 2014 11:01:13 +0000 (13:01 +0200)]
fixed relative path and prefix for adagrad-optimised autoencoder snapshots
Evan Shelhamer [Tue, 2 Sep 2014 03:05:57 +0000 (20:05 -0700)]
Merge pull request #1023 from longjon/unbreak-pycaffe
Fix pycaffe break due to dropping {blob,layer}.name
Jonathan L Long [Tue, 2 Sep 2014 02:32:24 +0000 (19:32 -0700)]
[pycaffe] use _blob_names, _layer_names instead of removed .name
Jonathan L Long [Tue, 2 Sep 2014 02:26:02 +0000 (19:26 -0700)]
[pycaffe] expose Net.blob_names and Net.layer_names
Jonathan L Long [Tue, 2 Sep 2014 02:36:49 +0000 (19:36 -0700)]
[pycaffe] add converter for vector<string> used by _*_names
Jeff Donahue [Mon, 1 Sep 2014 23:45:20 +0000 (16:45 -0700)]
add CUDA 6.5 error CUBLAS_STATUS_LICENSE_ERROR to cublasGetErrorString
enum
Jeff Donahue [Mon, 1 Sep 2014 21:51:59 +0000 (14:51 -0700)]
revert tools/train_net.cpp to previous, depecated version
Jeff Donahue [Mon, 1 Sep 2014 21:13:56 +0000 (14:13 -0700)]
Merge branch 'qipeng-solvers' into dev
Jeff Donahue [Mon, 1 Sep 2014 19:35:03 +0000 (12:35 -0700)]
make MNIST autoencoder solvers start from base_lr 0.01 and step (much
better performance) and terminate at iter 65K
Jeff Donahue [Mon, 1 Sep 2014 19:32:54 +0000 (12:32 -0700)]
make adagrad/nesterov train scripts follow new "run-from-root"
convention
Jeff Donahue [Mon, 1 Sep 2014 19:04:31 +0000 (12:04 -0700)]
Add "test-on-train" stage to test accuracy on the training data; correct
test_iter (should be 100 instead of 50)
Jeff Donahue [Mon, 1 Sep 2014 18:39:39 +0000 (11:39 -0700)]
mnist_autoencoder: always compute both cross-entropy loss and L2
(euclidean) error
qipeng [Tue, 26 Aug 2014 19:21:06 +0000 (12:21 -0700)]
hot fix for warning
qipeng [Tue, 26 Aug 2014 19:02:20 +0000 (12:02 -0700)]
lint
qipeng [Tue, 26 Aug 2014 19:01:26 +0000 (12:01 -0700)]
Re-added solver switch into the new caffe main excutable; fixed AdaGrad MNIST example
qipeng [Tue, 26 Aug 2014 06:17:51 +0000 (23:17 -0700)]
lint
qipeng [Tue, 26 Aug 2014 06:02:56 +0000 (23:02 -0700)]
Added sanity check for AdaGradSolver; added MNIST examples for solvers
Jeff Donahue [Fri, 22 Aug 2014 18:51:16 +0000 (11:51 -0700)]
Merge Test{SGD,AdaGrad,Nesterov}Solver; they become subclasses of
TestGradientBasedSolver
Jeff Donahue [Fri, 22 Aug 2014 18:51:29 +0000 (11:51 -0700)]
cleanup caffe.proto
qipeng [Thu, 21 Aug 2014 03:07:53 +0000 (20:07 -0700)]
added unit test for solvers and fixed solver bugs
qipeng [Wed, 20 Aug 2014 22:19:30 +0000 (15:19 -0700)]
proto conflit, lint, and math_functions (compiler complaint)
qipeng [Wed, 30 Jul 2014 02:56:36 +0000 (19:56 -0700)]
fixes after rebase
qipeng [Wed, 30 Jul 2014 02:46:37 +0000 (19:46 -0700)]
Addressed Yangqing's comments
qipeng [Tue, 29 Jul 2014 17:06:32 +0000 (10:06 -0700)]
fixed caffe.proto after a mistaken rebase
qipeng [Thu, 24 Jul 2014 20:09:28 +0000 (13:09 -0700)]
Added L1 regularization support for the weights
qipeng [Wed, 23 Jul 2014 17:25:44 +0000 (10:25 -0700)]
bugfixes for AdaGrad
qipeng [Wed, 23 Jul 2014 04:17:19 +0000 (21:17 -0700)]
improved numerical stability for AdaGrad
qipeng [Mon, 21 Jul 2014 17:47:55 +0000 (10:47 -0700)]
fixed solver constructor in train_net.cpp
qipeng [Sun, 20 Jul 2014 15:57:33 +0000 (08:57 -0700)]
converted pointers to shared_ptr
qipeng [Sun, 20 Jul 2014 00:33:35 +0000 (17:33 -0700)]
restored vituals in solver.hpp
qipeng [Sat, 19 Jul 2014 21:14:21 +0000 (14:14 -0700)]
Solver switching support & implementation of Nesterov's accelerated gradient and AdaGrad
Jeff Donahue [Mon, 1 Sep 2014 18:32:12 +0000 (11:32 -0700)]
use LMDB in mnist autoencoder examples
Evan Shelhamer [Mon, 1 Sep 2014 06:28:28 +0000 (23:28 -0700)]
make no GPU error in CPU-only mode a little clearer
Evan Shelhamer [Mon, 1 Sep 2014 06:25:31 +0000 (23:25 -0700)]
include comment on CPU mode fine-tuning for Flickr example
Evan Shelhamer [Mon, 1 Sep 2014 06:22:21 +0000 (23:22 -0700)]
Merge pull request #1014 from longjon/cleaner-pycaffe
Clean up pycaffe core
Evan Shelhamer [Sun, 31 Aug 2014 18:02:45 +0000 (11:02 -0700)]
Merge pull request #1008 from mohomran/mnist_with_lmdb
convert MNIST demo to lmdb, fixes
Evan Shelhamer [Sun, 31 Aug 2014 18:00:48 +0000 (11:00 -0700)]
Merge pull request #1015 from mohomran/fixing_parse_log_script
fix for tools/extra/parse_log.sh
Mohamed Omran [Sun, 31 Aug 2014 13:00:09 +0000 (15:00 +0200)]
parse_log.sh adapted to new training log format + fixed typos and updated description
Mohamed Omran [Sun, 31 Aug 2014 12:27:33 +0000 (14:27 +0200)]
updated lenet_train_test.prototxt + minor correction to create_mnist.sh
Mohamed Omran [Fri, 29 Aug 2014 15:16:41 +0000 (17:16 +0200)]
minor changes to variable names and error messages + set default backed in convert_mnist_data.cpp to lmdb
Mohamed Omran [Fri, 29 Aug 2014 15:04:25 +0000 (17:04 +0200)]
data now written to backend in batches
Mohamed Omran [Fri, 29 Aug 2014 14:04:24 +0000 (16:04 +0200)]
mnist demo now works with lmdb and leveldb (set parameter in create_mnist.sh), switched around includes
Jonathan L Long [Sun, 31 Aug 2014 08:14:58 +0000 (01:14 -0700)]
[pycaffe] expose Blob.Reshape
While this function should never be called in the current interface, it
will be necessary to have when embedding Python.
Jonathan L Long [Sun, 31 Aug 2014 07:45:06 +0000 (00:45 -0700)]
[pycaffe] remove name property from PyBlob and PyLayer
Attaching names to pycaffe Blobs and Layers isn't necessary anymore,
since they are now accessed by name. This change brings PyBlob and
PyLayer closer to their C++ counterparts, and makes it easier to pass
around blobs without knowing their names, e.g., when embedding Python
in a layer.
Jonathan L Long [Sun, 31 Aug 2014 07:36:26 +0000 (00:36 -0700)]
[pycaffe] declare the _caffe module init function
Jonathan L Long [Sun, 31 Aug 2014 07:26:12 +0000 (00:26 -0700)]
[pycaffe] split _caffe into source and header files
This brings pycaffe more in line with usual C++ style, and will allow
the wrapper classes to be used within caffe when embedding Python.
Jonathan L Long [Sun, 31 Aug 2014 07:07:43 +0000 (00:07 -0700)]
[pycaffe] make PyBlob a template over Dtype
We will not actually support double nets in Python with this commit, but
this change will be necessary to create layers that embed Python.
Jonathan L Long [Sun, 31 Aug 2014 07:01:20 +0000 (00:01 -0700)]
[pycaffe] make PyNet a class, not a struct
Jonathan L Long [Sun, 31 Aug 2014 06:55:57 +0000 (23:55 -0700)]
[pycaffe] use a namespace alias instead of using directives for boost::python
Jonathan L Long [Sun, 31 Aug 2014 06:49:44 +0000 (23:49 -0700)]
[pycaffe] live in caffe namespace instead of opening it
Jonathan L Long [Sun, 31 Aug 2014 06:48:41 +0000 (23:48 -0700)]
[pycaffe] use class names of the form Py* instead of Caffe*
Evan Shelhamer [Sun, 31 Aug 2014 04:37:12 +0000 (21:37 -0700)]
remove residual pthread references, but restore in build for gtest
gtest depends on pthread in Ubuntu 14.04
Evan Shelhamer [Sun, 31 Aug 2014 04:32:21 +0000 (21:32 -0700)]
ignore leveldb extension: ldb
Evan Shelhamer [Sat, 30 Aug 2014 23:59:21 +0000 (16:59 -0700)]
Merge pull request #1004 from kloudkl/ignore
Ignore leveldb more thoroughly
Evan Shelhamer [Sat, 30 Aug 2014 23:59:02 +0000 (16:59 -0700)]
fix up leveldb ignore
Evan Shelhamer [Sat, 30 Aug 2014 23:41:46 +0000 (16:41 -0700)]
Merge pull request #1010 from qipeng/boost-thread-with-nvcc
Fix boost-CUDA trouble in OS X build.
Evan Shelhamer [Sat, 30 Aug 2014 23:41:01 +0000 (16:41 -0700)]
Makefile: fix boost::thread linking, drop pthread, sort
Evan Shelhamer [Sat, 30 Aug 2014 23:37:51 +0000 (16:37 -0700)]
fix up renaming
qipeng [Sat, 30 Aug 2014 23:31:33 +0000 (16:31 -0700)]
renaming && typo fix
Evan Shelhamer [Sat, 30 Aug 2014 23:10:38 +0000 (16:10 -0700)]
Merge pull request #1012 from shelhamer/pr-policy
new PR policy: send master fixes + docs to master
Evan Shelhamer [Sat, 30 Aug 2014 23:06:17 +0000 (16:06 -0700)]
[docs] new PR policy: send master fixes + docs to master
To address fixes and documentation updates to master between official
releases from `dev` to `master` PRs will be accepted to `master`.
All new development must still branch off and PR to `dev`.
Jeff Donahue [Fri, 29 Aug 2014 10:31:33 +0000 (03:31 -0700)]
fix data_transformer param_name
Evan Shelhamer [Sat, 30 Aug 2014 22:37:37 +0000 (15:37 -0700)]
Merge pull request #1011 from YS-L/dev
Fix norm_region param in cifar10 deployment net
qipeng [Sat, 30 Aug 2014 17:38:33 +0000 (10:38 -0700)]
xcode compiler complaints (warnings)...
qipeng [Sat, 30 Aug 2014 17:28:15 +0000 (10:28 -0700)]
patch MacOS NVCC boost::thread issue
Yung Siang Liau [Sat, 30 Aug 2014 17:22:04 +0000 (01:22 +0800)]
Fix norm_region param in cifar10 deployment net
Evan Shelhamer [Sat, 30 Aug 2014 16:38:13 +0000 (09:38 -0700)]
back-merge
[docs] Several documentation fixes
Explain how to fix homebrew to allow updates after editing formulae
Evan Shelhamer [Sat, 30 Aug 2014 16:37:49 +0000 (09:37 -0700)]
Merge pull request #997 from ozancaglayn/dev
[docs] Several documentation fixes
Ozan Çağlayan [Thu, 28 Aug 2014 11:05:09 +0000 (14:05 +0300)]
[docs] Several documentation fixes
- Fix a typo and a path in MNIST tutorial
- Rework instructions about Fedora installation
Evan Shelhamer [Sat, 30 Aug 2014 16:35:07 +0000 (09:35 -0700)]
Merge pull request #1001 from drdan14/homebrew-git-repair
Explain how to fix homebrew to allow updates after editing formulae
Daniel Golden [Wed, 27 Aug 2014 16:25:10 +0000 (09:25 -0700)]
Explain how to fix homebrew to allow updates after editing formulae
Kai Li [Fri, 29 Aug 2014 06:14:32 +0000 (14:14 +0800)]
Ignore LevelDB files
Evan Shelhamer [Fri, 29 Aug 2014 05:40:59 +0000 (22:40 -0700)]
Merge pull request #970 from sergeyk/dev
Fine-tuning tutorial: adapt CaffeNet to Flickr Style data
Evan Shelhamer [Fri, 29 Aug 2014 04:45:46 +0000 (21:45 -0700)]
[example] edit fine-tuning and train on ~2000 images, 1557 / 382 split
- further detail merits of fine-tuning: less starving for itme and data
- set random seed for reproducing the tutorial
- 1557 train / 382 test split is more indicative of training quality
than splits of 200 images
Sergey Karayev [Sat, 23 Aug 2014 02:30:07 +0000 (19:30 -0700)]
[example] finetuning CaffeNet on Flickr Style data
Evan Shelhamer [Fri, 29 Aug 2014 02:47:53 +0000 (19:47 -0700)]
Merge pull request #1003 from shelhamer/examples-from-root
Standardize examples
Evan Shelhamer [Thu, 28 Aug 2014 23:34:57 +0000 (16:34 -0700)]
set examples paths relative to root
Evan Shelhamer [Thu, 28 Aug 2014 23:12:01 +0000 (16:12 -0700)]
ignore caffe generated files and stop ignoring examples
- ignore models and snapshots, means, and input databases
- excluding exceptions to commits is easier than including exhaustively
Evan Shelhamer [Thu, 28 Aug 2014 22:07:11 +0000 (15:07 -0700)]
Merge pull request #1000 from kloudkl/boost-thread
Replace pthread with boost::thread
Evan Shelhamer [Thu, 28 Aug 2014 17:55:33 +0000 (10:55 -0700)]
Back-merge to dev for doc fixes + cherry-picks
Back-merge these doc fixes:
clarify project origin
[docs] fix citation bibtex
All other changes are spurious commits from cherry-picking.
Kai Li [Thu, 28 Aug 2014 16:16:13 +0000 (00:16 +0800)]
The return value of WaitForInternalThreadToExit has reversed
Kai Li [Thu, 28 Aug 2014 16:05:49 +0000 (00:05 +0800)]
Add boost thread in the travis install script
Kai Li [Thu, 28 Aug 2014 11:17:07 +0000 (19:17 +0800)]
Replace pthread with boost::thread
Ozan Çağlayan [Mon, 25 Aug 2014 12:16:34 +0000 (15:16 +0300)]
[docs] Update installation docs to include Fedora
- All the instructions mentioned for CentOS/RHEL are also valid for Fedora Linux so add it.
- The package name for atlas is atlas-devel not libatlas-devel for CentOS/RHEL/Fedora so fix it.
- Add 'sudo' prefix to missing places to be consistent with the rest of the document.
- Add instructions for installing python headers in Fedora into the python support section.
- Update Linux installation section to reflect Fedora, refactor a bit.
Evan Shelhamer [Tue, 26 Aug 2014 16:51:19 +0000 (09:51 -0700)]
clarify project origin
Evan Shelhamer [Thu, 28 Aug 2014 04:43:53 +0000 (21:43 -0700)]
Merge pull request #857 from netheril96/gflags
Use gflags to parse command line arguments for convert_imageset
netheril96 [Thu, 28 Aug 2014 03:39:28 +0000 (11:39 +0800)]
create_imagenet.sh updated to new syntax
Jeff Donahue [Wed, 27 Aug 2014 23:01:23 +0000 (16:01 -0700)]
Merge pull request #993 from Yangqing/sweep
fix layer_factory.cpp bug: there should be no ifdefs
Yangqing Jia [Wed, 27 Aug 2014 22:40:14 +0000 (15:40 -0700)]
fix layer_factory.cpp bug: there should be no ifdefs
Evan Shelhamer [Wed, 27 Aug 2014 07:18:09 +0000 (00:18 -0700)]
Merge pull request #977 from ozancaglayan/patch-1
[docs] Update installation docs to include Fedora
Evan Shelhamer [Tue, 26 Aug 2014 23:26:15 +0000 (16:26 -0700)]
Merge pull request #984 from shelhamer/drop-curand-reset
Drop obsolete CURAND reset for CUDA 6.5 compatibility
Evan Shelhamer [Tue, 26 Aug 2014 21:32:14 +0000 (14:32 -0700)]
default ilsvrc solving to GPU