[tests] Model tests expansion (partially undone in #1464)

[ourownstory]

to date, the test horizon of our model benchmarks has been rather long (20%), yielding in accuracy over a long generalization horizon being tested. This skews results towards more regularized models.

This PR introduces dual test horizons - a short one (5-10%) and a long one (20-30%).
Those will cover regular and long-horizon model fit accuracy, respectively.

Further, the 3 tutorial datasets (3 variations of the kaggle energy data) are introduced to be used in future tests.

Note: The Accuracy of the tests is currently lower due to early stopping having been disabled in this version. This is however necessary as early stopping produces more irregular results with hard to interpret training graphs as part of training is missing. This is particularly evident for AirPassengers.

[codecov]

Codecov Report

Merging #1455 (d4d8460) into main (a22a579) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main    #1455   +/-   ##
=======================================
  Coverage   88.32%   88.32%           
=======================================
  Files          38       38           
  Lines        5097     5097           
=======================================
  Hits         4502     4502           
  Misses        595      595           

[github-actions]

Model Benchmark

Benchmark	Metric	main	current	diff
AirPassengers_test30	MAE_val	-	25.3608	-
AirPassengers_test30	RMSE_val	-	30.624	-
AirPassengers_test30	Loss_val	-	0.0074	-
AirPassengers_test30	MAE	-	5.3982	-
AirPassengers_test30	RMSE	-	6.8828	-
AirPassengers_test30	Loss	-	0.0002	-
AirPassengers_test30	time	-	19.57	-
YosemiteTemps	MAE_val	1.34899	0.44287	-67.17%	🎉
YosemiteTemps	RMSE_val	2.00817	0.62874	-68.69%	🎉
YosemiteTemps	Loss_val	0.00078	7e-05	-91.32%	🎉
YosemiteTemps	MAE	1.32133	0.80138	-39.35%	🎉
YosemiteTemps	RMSE	2.13713	1.44081	-32.58%	🎉
YosemiteTemps	Loss	0.00064	0.00027	-57.67%	🎉
YosemiteTemps	time	59.72889	138.28	131.51%	❌
PeytonManning	MAE_val	0.58162	0.34778	-40.21%	🎉
PeytonManning	RMSE_val	0.72218	0.49928	-30.86%	🎉
PeytonManning	Loss_val	0.01239	0.00595	-51.96%	🎉
PeytonManning	MAE	0.41671	0.35123	-15.71%	🎉
PeytonManning	RMSE	0.55961	0.48066	-14.11%	🎉
PeytonManning	Loss	0.00612	0.0046	-24.8%	🎉
PeytonManning	time	12.73383	49.71	290.38%	❌
AirPassengers	MAE_val	13.0627	30.1272	130.64%	❌
AirPassengers	RMSE_val	15.94532	31.014	94.5%	❌
AirPassengers	Loss_val	0.00131	0.00371	182.63%	❌
AirPassengers	MAE	9.88153	6.27639	-36.48%	🎉
AirPassengers	RMSE	11.73543	7.88805	-32.78%	🎉
AirPassengers	Loss	0.00052	0.00019	-62.92%	🎉
AirPassengers	time	5.30237	20.39	284.54%	❌
YosemiteTemps_test30	MAE_val	-	1.8	-
YosemiteTemps_test30	RMSE_val	-	2.2687	-
YosemiteTemps_test30	Loss_val	-	0.001	-
YosemiteTemps_test30	MAE	-	0.8432	-
YosemiteTemps_test30	RMSE	-	1.4617	-
YosemiteTemps_test30	Loss	-	0.0003	-
YosemiteTemps_test30	time	-	119.61	-
PeytonManning_test30	MAE_val	-	0.9731	-
PeytonManning_test30	RMSE_val	-	1.128	-
PeytonManning_test30	Loss_val	-	0.0317	-
PeytonManning_test30	MAE	-	0.33	-
PeytonManning_test30	RMSE	-	0.4574	-
PeytonManning_test30	Loss	-	0.0044	-
PeytonManning_test30	time	-	43	-

Model training plots

Model Training

PeytonManning

YosemiteTemps

AirPassengers