Python notebook for computing very simple classification baselines.
Improved fabfile modularization and AWS feature support.
Remove type hints from LSTM model code for AWS compatibility. :(
Experimental tool (sanity.py) for:
a) Ensuring downloaded checkpoints are evaluated correctly locally (i.e. that the data you have locally is the same used remotely for training, and that your validation results for submitting to Kaggle are produced using the correct preprocessing).
b) Producing very simple ensemble predictions.
sanity.py) for: a) Ensuring downloaded checkpoints are evaluated correctly locally (i.e. that the data you have locally is the same used remotely for training, and that your validation results for submitting to Kaggle are produced using the correct preprocessing). b) Producing very simple ensemble predictions.