-
-
Notifications
You must be signed in to change notification settings - Fork 616
Dense with optional reshape #856
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
Conversation
|
This is interesting! We should also add a depwarn to the changed API. |
|
Glad someone likes I can add a news line, but should it go under v0.10.0? |
src/layers/basic.jl
Outdated
|
|
||
| """ | ||
| Dense(in::Integer, out::Integer, σ = identity) | ||
| Dense(in => out, σ = identity) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Shouldn't we have in this docstring, or as separete docstrings, also the other constructors?
Dense(in, out)
Dense(W, b)There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
See if you like this now. I’ve also tried to make clearer that σ is not a keyword.
|
What is the status of this? |
|
Would still be nice, IMO, but didn't gain traction? I guess an alternative would be to make a nice Instead of |
|
+1 for the Reshape layer, to me it seems more manageable and of broader utility |
|
+1 for the Reshape layer as well |
Fixes #282, by allowing
Dense((5, 3), (2, 7))as suggested. Also closes #293, #617.For
d = Dense(15, 14)I wanted to leave the original behaviour alone, but attempting to do this by dispatch landed me in ambiguity hell, so it's anifstatement (which I think ought to get compiled out). Possibly this should have&& ndims(x)<=2so that suchddoes allow higher-dimensional data.I also added a line to allow
Dense(15 => 14)andDense((5, 3) => (2, 7)), more likeConvand less confusing I think. But will remove if anyone objects / thinks this deserves its own issue.Note by the way the following
::Anyproblem. Since this persists after] free Flux, I don't think it's my fault. (This is on Julia 1.1)