Conv layer: fix crash by setting weight pointer #1048
Merged
Conversation
This file contains 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
|
Eep, nice catch, and thanks for the fix! |
jeffdonahue
added a commit
that referenced
this pull request
Sep 8, 2014
Conv layer: fix crash by setting weight pointer
mitmul
pushed a commit
to mitmul/caffe
that referenced
this pull request
Sep 30, 2014
Conv layer: fix crash by setting weight pointer
shelhamer
added a commit
that referenced
this pull request
Oct 1, 2014
fix NULL pointer when back-propagating to the data for fixed i.e. not back-propagated weights. same issue as #1048
|
Fixed for CuDNNConvolutionLayer in a02ec65. |
mitmul
pushed a commit
to mitmul/caffe
that referenced
this pull request
Oct 11, 2014
fix NULL pointer when back-propagating to the data for fixed i.e. not back-propagated weights. same issue as BVLC#1048
RazvanRanca
pushed a commit
to RazvanRanca/caffe
that referenced
this pull request
Nov 4, 2014
Conv layer: fix crash by setting weight pointer
RazvanRanca
pushed a commit
to RazvanRanca/caffe
that referenced
this pull request
Nov 4, 2014
fix NULL pointer when back-propagating to the data for fixed i.e. not back-propagated weights. same issue as BVLC#1048
cbfinn
pushed a commit
to cbfinn/caffe
that referenced
this pull request
Feb 26, 2015
fix NULL pointer when back-propagating to the data for fixed i.e. not back-propagated weights. same issue as BVLC#1048
slayton58
pushed a commit
to slayton58/caffe
that referenced
this pull request
Mar 4, 2015
fix NULL pointer when back-propagating to the data for fixed i.e. not back-propagated weights. same issue as BVLC#1048
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
The problem is that in ConvolutionLayer::Backward_[gpu,cpu], the pointer "weight" is set only if the this->param_propagate_down_[0] is true, otherwise it is left NULL. However "weight" is used if either this->param_propagate_down_[0] or propagate_down[i] is true. So in the case where the first bool is false but the second is true, we end up adding weight offset to a NULL "weight". This caused a seg fault running on a CPU and cuda error 13 running on GPU.
I hit this in a scenario where I was doing layerwise training of a convolutional autoencoder. The learning rate for the already-trained layers was set to zero, and the newly-added layers had non-zero learning rate.