Second Training

Second training, with the following parameters:

Parameter	Value
num_samples	2000
epochs	256
learning_rate	1e-3
batch_size	128
validation_split	0.66

Initial Training

First model release, trained with 1000 samples from each letter and with the following parameters:

EPOCHS = 64
LR = 1e-3
BS = 128