The ScatterLog
log.append(x=x, y=y, color=c)
function has a color argument. However, we cannot actually pass a colour to it, it just seems to take whatever value you give it, and colors all points with the same color value the same, according to some colormap.
It is a little bit counter intuitive, and I think it would make it much more useful if we could actually e.g. pass it a scalar value and use the scalar value along with some colormap, or pass an explicit color per point - a little bit more in the sense of matoplotlib.pyplot.scatter(..., c=c). This would then allow us to use c as an third dimension, not only for discrete values, but also for continuous values, e.g., a loss or metric.
The ScatterLog
function has a
colorargument. However, we cannot actually pass a colour to it, it just seems to take whatever value you give it, and colors all points with the same color value the same, according to some colormap.It is a little bit counter intuitive, and I think it would make it much more useful if we could actually e.g. pass it a scalar value and use the scalar value along with some colormap, or pass an explicit color per point - a little bit more in the sense of
matoplotlib.pyplot.scatter(..., c=c). This would then allow us to usecas an third dimension, not only for discrete values, but also for continuous values, e.g., a loss or metric.