Right now you can't use "spearman" and "pearson" metrics when training multi-task: "#41
The metrics API is really finicky and probably has other issues as well. I'm not sure it makes sense to continue utilizing it despite it's conveniences.
Worth a brainstorm to figure out the best way to approach, and will likely be connected to the SeqModels revamp
Right now you can't use "spearman" and "pearson" metrics when training multi-task: "#41
The metrics API is really finicky and probably has other issues as well. I'm not sure it makes sense to continue utilizing it despite it's conveniences.
Worth a brainstorm to figure out the best way to approach, and will likely be connected to the SeqModels revamp