README and cleanup

(minor refactor) os.path to pathlib and tdqm to infer.py

(debug) fix eval loop not selecting randomly

Save test visuals during final eval

Merge branch 'main' of https://github.com/MRiabov/wire-seg-hr-impl

(debug) to torch operations in bfloat conversions.

Add final checkpoint save and test set val

Eval to infer.py

(optim) benchmark=True and higher eval batch size

Speeding up eval by reducing iters

Added precision training option

(optim) prefetch and speed up eval.

(debug) 6 chan instead of 7. (warn: conflicting git changes)

to mit_b2 as per paper

(debug) eval fixes

(devops) automatic pull and preprocess of datasets

Running training