models from first round of training + some files to enable predictions to be to be run as SLURM jobs

added functionality to model creation that allows for learned embeddings (or combo of learned and hardcoded). Also, adjusted amino_acid.py to make X's index = 0 so that it can be used as a mask.

altered training init to work with SLURM and corrected typo (BLOSUM --> BLOSUM62) in hyperparameters, specifically encoding options

fixed nonsensical import of subpackage in middle of script to point to mhc_rank instead of mhcflurry

trying to figure out why cleave_radius is an unexpected keyword arg in vector_encode even though vector encode has an explicit argument cleave_radisu...

Fixed big in argparse for make_architectures.py script. Created file with architecture options (1296) and saved as hyperparameters.json to ap_models dir

debugged architecture / training & created script to make models fitting all combinations of varied hyperparameters for tuning

added feature to flank_encoding.py, enabling pssm scores to be added as first position in residue encoding vectors