Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fix bug in recurrent branch #6

Open
wants to merge 26 commits into
base: master
Choose a base branch
from
Open

Fix bug in recurrent branch #6

wants to merge 26 commits into from

Conversation

cvondrick
Copy link

longjon and others added 26 commits March 13, 2015 13:12
@longjon @jeffdonahue
zero-init param diffs and accumulate gradients
6965746 
(With layers whose backwards accumlate gradients), this effectively
decouples the computational batch from the SGD minibatch. Each
iteration accumulates gradients over iter_size batches, then parameters
are updated.
@longjon @jeffdonahue
zero-init param diffs in gradient checker
37bddae
@sguada @jeffdonahue
accumulate gradients in inner product layer
05185e2
@longjon @jeffdonahue
accumulate gradients in (de)conv layers
c8f0bbe
@longjon @jeffdonahue
accumulate gradients in cudnn conv layer
2983572
@jeffdonahue
Add gpu_util.cuh, with caffe_gpu_atomic_add
ea39cb5 
(double impl from NVIDIA dev docs; float impl included in CUDA as
"atomicAdd")
@jeffdonahue
test_gradient_check_util: check_bottom < -1 only checks params
0c19b6b
@jeffdonahue
Add EmbedLayer for inner products with sparse input (one-hot vectors),
bb8a91d 
with unit tests
@jeffdonahue
EmbedBackward with no loops -- use caffe_gpu_atomic_add instead
9271abf
@jeffdonahue
Add (very simple version of) ReshapeLayer
15c71c4
@jeffdonahue
FlattenLayer fix -- top should always Share* from bottom (and do
b6fd6b2 
everything in Reshape)
@jeffdonahue
AccuracyLayer: add 'denominator' param
af0ded7
@jeffdonahue
EltwiseLayer can take a blob of per-num coefficients
d3e4e30
@jeffdonahue
EltwiseLayer with coeff blob GPU kernel
cb4d64f
@jeffdonahue
Allow SliceLayer to have a single top Blob (for testing)
5066741
@jeffdonahue
Allow ConcatLayer to take a single bottom Blob (for testing)
0f4afa0
@jeffdonahue
Modifications to Net to facilitate unrolled recurrent networks
6cda560
@jeffdonahue
TestNet fixes for Net weight sharing modifications
9eca769
@jeffdonahue
Add RecurrentLayer: an abstract superclass for other recurrent layer …
899d345 
…types
@jeffdonahue
Add RNNLayer, with tests
579162e
@jeffdonahue
Add LSTMLayer and LSTMUnitLayer, with tests
97284a1
@jeffdonahue
Add scripts for downloading COCO2014 tools & data
555a5b4
@jeffdonahue
Add scripts to create HDF5 datasets from COCO captions
89a1ae4
@jeffdonahue
Prototxts + script for training COCO caption language model
dfb4444
@jeffdonahue
Prototxts + script for training LRCN COCO image captioning model
34230c6
fixing reshape
3e1a0ff
@dxj19831029
Copy link

I think your target should be recurrent branch?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
5 participants
@cvondrick @dxj19831029 @longjon @sguada @jeffdonahue