Expose Net.copy_trained_layers_from and Net.share_trained_layers_with in pycaffe#1195
Merged
Expose Net.copy_trained_layers_from and Net.share_trained_layers_with in pycaffe#1195
Conversation
longjon
changed the title
Member
|
I like the suggested name change to |
This allows finetuning from Python, for example.
longjon
force-pushed
the
python-copy-from
branch
from
f2d5930 to
0dfda37
Compare
shelhamer
added a commit
that referenced
this pull request
Oct 10, 2014
Expose Net.copy_from and Net.share_with in pycaffe
mitmul
pushed a commit
to mitmul/caffe
that referenced
this pull request
Oct 11, 2014
Expose Net.copy_from and Net.share_with in pycaffe
RazvanRanca
pushed a commit
to RazvanRanca/caffe
that referenced
this pull request
Nov 4, 2014
Expose Net.copy_from and Net.share_with in pycaffe
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 allows finetuning from Python, for example.
Misgivings:
copy_trained_layers_fromandshare_trained_layers_withare very long names, while pycaffe mostly has pretty succinct naming. Also, the layers don't really have to be trained at all, so maybe the simplecopy_fromandshare_withwould be better. But these are the names in the C++ interface.I don't know what will happen if you
share_trained_layers_witha net, and then delete that net. But it's the same as what will happen if you do the same in C++; does anyone know how that is handled? For the use case I have in mind (the solver), this issue does not arise, because the solver is holding all the nets.