Print DVC plots and metrics directly in the terminal

DVC has changed the way I handle Machine Learning projects for the better. The dvc plots diff command is a great way to see the effect of rounds of experimentation.

If you use the kitty terminal, it's possible to print DVC plots directly in the terminal without opening an external program/browser. I wrote a function for the fish shell to simplify this process.

Assuming you are using fish as your shell, save this code as ~/.config/fish/functions/plotit.fish

	# Usage: plotit <target (required)> <x> <y> <width> <height>
	function plotit -d="Prints DVC plots on the terminal" -a target x y width height
	if test -z "$target"
	echo Please provide a target csv file.
	return 1
	end
	# set `-x` or `-y` if they are provided
	set x_or $x ""
	set y_or $y ""
	if test $x_or[1] != ""
	set x_or -x $x_or
	end
	if test $y_or[1] != ""
	set y_or -y $y_or
	end
	# set fallback width and height
	test -z "$width"; and set width 500
	test -z "$height"; and set height 500
	set dvcplotdiff dvc plots diff $x_or $y_or --show-vega --targets $target
	eval $dvcplotdiff | \
	jq ". += {\"width\": $width, \"height\": $height}" | \
	vl2vg | \
	vg2svg | \
	kitty +kitten icat --align left && \
	dvc metrics diff
	end

view raw plotit.fish hosted with ❤ by GitHub

Then refresh your shell with source ~/.config/fish/functions/plotit.fish. Finally, install jq and vega-cli, if you don't have them in your system yet.

Example usage when inside a DVC-enabled project:

• plotit pr_curve.csv -> prints diff plot with width x height of 500px x 500px and x set to row-index and y to the second column of the csv
• plotit pr_curve.csv recall precision -> same as above but selects the columns for x and y
• plotit pr_curve.csv recall precision 800 800 -> with custom width and height in pixels

Enjoy!