Fixes #18:
Wrote unit tests, asserted implementation was incorrect, and fixed it by small modifications:
model.eval() in feature_extraction module (switches model state from train to eval)
store intermediate results in a list rather than in a concatenated tensor (maintain order & presumably more efficient)
Fixes #18: Wrote unit tests, asserted implementation was incorrect, and fixed it by small modifications: model.eval() in feature_extraction module (switches model state from train to eval) store intermediate results in a list rather than in a concatenated tensor (maintain order & presumably more efficient)