fix broken tests and fix v0.6 deprecations#3
Conversation
| # input and target are pre-populated... compute the output value as: ∑ loss(targetᵢ, inputᵢ) | ||
| function transform!(lt::LossTransform) | ||
| lt.output.val[1] = sumvalue(lt.loss, value(lt.target), input_value(lt)) | ||
| lt.output.val[1] = value(lt.loss, value(lt.target), input_value(lt), AvgMode.Sum()) |
There was a problem hiding this comment.
I changed this based on JuliaML/LossFunctions.jl@774014d#diff-f2fc3674e12bf71a707cb4988076d39eR9
|
I also updated Travis to add v0.6 and remove osx testing (OSX testing is slow and, IMO, not super useful for pure-julia packages like this). |
Evizero
left a comment
Evizero
left a comment
There was a problem hiding this comment.
lgtm. We can reactive OSX when the package is more stable.
|
I'll merge once CI passes. Thanks! |
REQUIRE
Outdated
| LearnBase | ||
| LossFunctions | ||
| PenaltyFunctions | ||
| Transformations |
There was a problem hiding this comment.
Is Transformations in METADATA? I doubt it. We may need to put a Pkg.clone in .travis.yml here https://github.com/JuliaML/ObjectiveFunctions.jl/blob/master/.travis.yml#L14 instead
There was a problem hiding this comment.
Yup, just realized that.
|
I am trapped in a nightmare world of cyclic dependencies between this package, StochasticOptimization, and Transformations 😦 |
|
Also, PenaltyFunctions is 0.6-only, so I'm bumping up the REQUIRE and travis versions |
|
Next blocker: JuliaML/Transformations.jl#5 |
|
Sorry that this is frustrating. Know that your effort is appreciated |
|
Hey, it's still better than my research code 😉 |
|
I also have high hopes that Pkg3 will make managing this kind of distributed ecosystem much easier, since a lot of this trouble could be avoided if it were easier for us to clearly specify which versions of each of these packages worked well together. As it is, I've had to use some hack-y tools to manage packages pointing to specific git SHAs in my own work. |
|
Woohoo! One down! |
|
Thanks! |
2 participants
No description provided.