summaryrefslogtreecommitdiff
path: root/src/caffe
AgeCommit message (Collapse)AuthorFilesLines
2015-01-25Merge pull request #1755 from jeffdonahue/softmax-optimizationEvan Shelhamer1-24/+24
SoftmaxLayer GPU optimization
2015-01-25Merge pull request #1754 from jeffdonahue/softmax-loss-fixJeff Donahue1-0/+4
SoftmaxWithLossLayer: use CreateLayer
2015-01-21hdf5_save_nd_dataset takes a const string& (instead of const string)Jeff Donahue1-2/+2
2015-01-21SoftmaxWithLossLayer: use CreateLayer so that a CuDNNSoftmaxLayerJeff Donahue1-0/+4
is created if available
2015-01-21Back-merge fixes + docsEvan Shelhamer1-7/+8
and other fixes and documentation updates.
2015-01-19Unroll kernels in SoftmaxLayer...from terrible performance to mediocreJeff Donahue1-24/+24
performance.
2015-01-19SetTotalBytesLimit to the max (2 GB minus 1 byte)Jeff Donahue1-1/+3
2015-01-17Make comments for sparse GaussianFiller match actual behaviorJeff Donahue1-1/+1
(Fixes #1497 reported by @denizyuret)
2015-01-16Merge pull request #1686 from longjon/net-constJon Long2-18/+21
Improve const-ness of Net
2015-01-15Merge pull request #1236 from mlapin/legacy_nvcc_supportEvan Shelhamer4-4/+32
Drop OpenCV includes from NVCC code for legacy reasons.
2015-01-15lint internal threadEvan Shelhamer1-1/+2
2015-01-15Merge pull request #1335 from ryotat/masterEvan Shelhamer1-7/+7
Fix leaking thread and groom internal thread implementation.
2015-01-10fix typo in layer_factory.cppJonathan L Long1-1/+1
2015-01-09improve const-ness of NetJonathan L Long2-18/+21
2015-01-02break out Step from SolverJonathan L Long1-35/+38
2014-12-30[tests] don't use Gaussian labels in NetTest's TinyNetJonathan L Long1-0/+4
Previously, labels were being nonsensically filled with a zero-mean Gaussian. Now they are just set to constant zero. Note that this is not necessary for TrickyNet, neither forward nor backward are ever called.
2014-12-29remove SoftmaxLossLayer CPU_ONLY stubs, since there is no GPU versionJonathan L Long1-5/+1
2014-12-29move softmax loss GPU todo comment to header fileJonathan L Long1-27/+0
There's no need to have GPU stubs just for a comment, and the absence of a GPU-specific implementation is made more obvious by the nonexistence of the .cu file.
2014-12-27update use of GetLayer -> LayerRegistry::CreateLayerJonathan L Long1-1/+2
2014-12-22Warning of fallback only the first time for cudnn_pooling_layerSergio1-2/+4
2014-12-21Fix init current_stepSergio1-0/+1
2014-12-21Added Multistep, Poly and Sigmoid learning rate decay policiesSergio2-8/+37
Conflicts: include/caffe/solver.hpp src/caffe/proto/caffe.proto src/caffe/solver.cpp
2014-12-21Display averaged loss over the last several iterationsqipeng2-2/+20
2014-12-20Add CHECKs to prevent segfault for incorrect IMAGE_DATA layers.Thomas Schenker1-0/+5
Incomplete IMAGE_DATA layers cause a segmentation fault, e.g.: layers { name: "mnist" type: IMAGE_DATA top: "data" top: "label" image_data_param { batch_size: 100 } include: { phase: TRAIN } } or layers { name: "mnist" type: IMAGE_DATA top: "data" top: "label" data_param { source: "./data/train.txt" batch_size: 100 } include: { phase: TRAIN } } Also empty image list files cause a segfault. This is fixed by adding 3 CHECK instructions.
2014-12-10Check input line count in HDF5 data layerJonathan Ho1-0/+2
2014-12-05compile for compute capability 5.0Evan Shelhamer1-1/+3
2014-11-30remove redundant code in ConvolutionLayer::ReshapeJonathan L Long1-3/+0
2014-11-25Merge pull request #1455 from seanbell/tanh-fixJeff Donahue3-5/+103
Fix tanh layer for badly scaled input values
2014-11-23use DCHECK in SoftmaxLossLayer so that bounds checking has no perf penalty ↵Jonathan L Long1-2/+2
without DEBUG
2014-11-22in SoftmaxLossLayer, check label >= 0 in addition to upper boundJonathan L Long1-0/+1
2014-11-19Fixed header order to satisfy linterSean Bell1-2/+1
2014-11-19fixed tanh to not return NaN for input values outside the range [-40, 40]Sean Bell3-5/+104
2014-11-13fix relu cudnn test bugYangqing Jia1-2/+5
2014-11-07relax benchmark milliseconds thresholdYangqing Jia1-4/+6
2014-11-07clean incorrect relu test codeYangqing Jia1-6/+14
2014-10-22Minor whitespace fix in logging message in HDF5 output layerMartin Baeuml1-1/+1
2014-10-21Forward declare boost::thread instead of including boost/thread.hpp in ↵Ryota Tomioka1-0/+6
internal_thread.hpp.
2014-10-20Merge pull request #1326 from jackculpepper/absgradatzeroEvan Shelhamer2-3/+2
[fix] define gradient at zero to be zero to fix ABSVAL NaN
2014-10-20Fixed a memory leak issue in InternalThread (and removed caffe::Thread calss).Ryota Tomioka1-7/+1
2014-10-19include opencv only in .cppMaksim Lapin4-4/+32
2014-10-19define gradient at zero to be zeroJack Culpepper2-3/+2
2014-10-18LMDB doesn't support many concurrent read-only transactions, so this ↵Kevin James Matzen2-43/+72
preallocates one read-only transaction and reuses it. It's very important that iterations are considered invalid after a commit has been performed.
2014-10-17back-mergeSergey Karayev1-1/+1
2014-10-16Fix init current_stepSergio1-0/+1
2014-10-15Merge pull request #190 from sguada/new_lr_policiesSergio Guadarrama2-8/+37
New lr policies, MultiStep and StepEarly
2014-10-15Added CPUTimerSergio4-28/+75
Make timing more precise using double and microseconds
2014-10-15Upgrade compute_image_mean to use gflags, accept list_of_images, and print ↵Sergio1-2/+2
mean_values
2014-10-15Change caffe time to do forward/backward and accumulate time per layerSergio4-60/+12
2014-10-15Added cache_images to WindowDataLayerSergio2-6/+36
Added root_folder to WindowDataLayer to locate images
2014-10-15Speed up WindowDataLayer and add mean_valuesSergio1-19/+46
generated by cgit v1.2.3 (git 2.25.1) at 2024-10-17 00:50:08 +0000