multi-label accuracy layer?

[Achal Dave]

I've read a few posts on here about multi-label classification (i.e. each of my images can have more than one label each), but I still can't seem to find the mythical multi-label accuracy layer. (https://groups.google.com/d/msg/caffe-users/6SnEj7p5KGk/yasbbJRzKQAJ has a decent summary of the relevant pull requests and issues, and a post that says "you need to implement your own accuracy layer or get it from a PR", but it's not clear what PR provides it).

Is there a canonical multi-label AccuracyLayer pull request that I'm missing? Is there a reason one doesn't exist in Caffe? If I write my own, should I try to submit a pull request?