Closed
Conversation
…nclude glog at some point. Including caffe/common.hpp. (2) I often misconfigure some layer and softmax breaks when the dimensionality is too small for the input layers. Check and fail fast. (3) CV_LOAD_IMAGE_COLOR is deprecated and has been removed in OpenCV 3.0. Replaced with cv::IMREAD_COLOR.
[fix] build with OpenCV 3 and other minor fixes
Clean flaky code
Minor change: some namespace simplification
Give back to layer what is layer's, and to factory what is factory's
compilation on working systems This reverts commit 4587b2f.
some more namespace cleaning.
…then they might fail.
…at was recently removed.
…r. Removed two buffer copies in dereference operation for DB iterators.
…atched what we expected.
… a few bugs related to ReadOnly mode in Database in order to pass test cases.
…t-increment operator so that it forks off a copy of the LevelDB or LMDB iterator/cursor when necessary. Neither of these APIs allow you to directly copy an iterator or cursor, so I create a new iterator and seek to the key that the previous one was currently on. This means the pre-increment operator can be much cheaper than the post-increment operator.
…onditions inside open, put, get, and commit, these functions return a bool indicating whether or not the operation was successful or a failure. This means the caller is now responsible for error checking.
…or put and key by const reference for put. Additional copies are made for get and put in the LMDB implementation.
…y having each iterator hold a shared pointer to the DB. I manually specified a deconstructor for the LeveldbState to make it clear what order these two things need to be deallocated in.
The gradient checker fails on certain elements of the PowerLayer checks, but only 1-3 sometimes fail out of the 120 elements tested. This is not due to any numerical issue in the PowerLayer, but the distribution of the random inputs for the checks. boost 1.56 switched the normal distribution RNG engine from Box-Muller to Ziggurat.
Fix PowerLayer gradient check failures by reducing step size
Die on inputs that fail to load
GPU version of SoftmaxWithLossLayer
Feed cv::Mats to MemoryDataLayer and set batch size on-the-fly.
uint64_t size_t
These layers are meant for generic inputs so they do not do transformations. As `transform_param` was pulled out of data layers' proto definitions for reuse, these layers silently ignored transformation parameters until now instead of signalling their refusal.
Bug in MemoryData that does not allow using arrays with n_ * size_ > 2^31
…t-fix Fix for CuDNN layer tests: only destroy handles if setup
HDF5_DATA + MEMORY_DATA refuse loudly to transform
Gradient clipping
BlobMathTest: fix precision issues
Introduced by Layer type is a string BVLC#1694
… for use with LRUs instead of tanh. Based on paper: He et al, "Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification," 2015
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR implements the weight initialization strategy from http://arxiv.org/abs/1502.01852. It is very similar to the Xavier filler---except designed for RLU instead of tanh non-linearities.
This would compliment #1880 which implements the PReLU layer also proposed in this paper.