Fix sequence size and compute_validation_tests

[isaacmg]

The goal of this PR is to do the following

• Correct a bug where forecast_length was not being passed to `simple_dec
• Fix bugs related to the unscaled_test_loss
• Add unit tests to validate compute_validation logic explicitly.

[codecov]

Codecov Report

❌ Patch coverage is 84.37500% with 10 lines in your changes missing coverage. Please review.
✅ Project coverage is 60.70%. Comparing base (7343cfb) to head (f6d48be).
⚠️ Report is 1625 commits behind head on masterr.

Files with missing lines	Patch %	Lines
flood_forecast/pytorch_training.py	61.53%	5 Missing ⚠️
flood_forecast/basic/linear_regression.py	50.00%	2 Missing ⚠️
tests/validation_tests.py	95.65%	2 Missing ⚠️
flood_forecast/custom/dilate_loss.py	0.00%	1 Missing ⚠️

❗ There is a different number of reports uploaded between BASE (7343cfb) and HEAD (f6d48be). Click for more details.

HEAD has 5 uploads less than BASE

Flag	BASE (7343cfb)	HEAD (f6d48be)
python	18	13

Additional details and impacted files

@@             Coverage Diff              @@
##           masterr     #264       +/-   ##
============================================
- Coverage    71.06%   60.70%   -10.37%     
============================================
  Files           56       50        -6     
  Lines         3543     3275      -268     
============================================
- Hits          2518     1988      -530     
- Misses        1025     1287      +262     

Flag	Coverage Δ
python	60.70% <84.37%> (-10.37%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[isaacmg]

This #265 #253

[isaacmg]

Update there is still an ongoing problem that I can't figure out where in the cases of unscaled_test_loss MSE seems to be double DilateLoss in pretty much every training session. This is try in every logging for some reason. This isn't the case for the scaled_test. At the moment I believe the scaled_test_loss is correct. Additionally when un-scaling the scaled_test_loss it often doesn't equal the scaled_test_loss.