Differences in add_task and add_transformer for PL base class

[david-z-shi]

My issue is about: seeing differences in the behavior of add_task and add_transformer methods implemented in the progressive_learner class. Calling add_task and add_transformer for different tasks lead to drastically different BTE behaviors, even though add_task just calls add_transformer.

Figure from @srahul1222.

• Package Version [0.0.4]

[david-z-shi]

@jdey4
@srahul1222 if you can provide a link to the notebook of the figure above

[srahul1222]

Link to notebook and functions file for new figure (using add_task for all tasks):

• https://github.com/srahul1222/progressive-learning/blob/flowers/docs/experiments/fte_bte_flowers.ipynb
• https://github.com/srahul1222/progressive-learning/blob/flowers/docs/experiments/functions/fte_bte_flowers_functions.py