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.

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

UI network design tool #60

Open
lukeyeager opened this issue Apr 10, 2015 · 4 comments
Open

UI network design tool #60

lukeyeager opened this issue Apr 10, 2015 · 4 comments

Comments

@lukeyeager
Copy link
Member

This has been mentioned many times, most recently at BVLC/caffe#481 (here).

Nice to have:

  1. Be able to select TRAIN, VAL and TEST phases for each layer, not just caffe's TRAIN and TEST. Related to Replace input fields with a proper layer BVLC/caffe#1245.
  2. Do this in a way that doesn't depend on caffe specifically, but which supports a list of layers from multiple backends. So if you choose caffe you have a certain list of layers and features to work with. If you choose another framework, the same UI tool should work.
@yankov
Copy link

yankov commented Apr 10, 2015

What frameworks should be supported besides caffe?

@lukeyeager
Copy link
Member Author

Torch and Theano would probably be the next frameworks added. Torch support is already in progress internally and we should be pushing out a branch with those changes before too long.

I wouldn't worry about actually supporting any Torch layers yet - we just want a design such that we can swap out the list of available layers and options at will. In other words, this should be abstracted away from caffe's NetParameter prototxt definition.

If you want to work on getting this added to DIGITS that would be great! If not, it would probably be a while before I got around to it.

@dzhwinter
Copy link

I wanna do that.

@lukeyeager lukeyeager added the UI label Jun 19, 2015
@man-sean
Copy link

Be able to select TRAIN, VAL and TEST phases for each layer, not just caffe's TRAIN and TEST. Related to BVLC/caffe#1245.

It will be nice if we could also specify weights for each layers from existing models

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

4 participants