def get_hyperparam_config(dataset):
c = initialise_transformer_arguments()
c['mode'] = 'train'
c['exp_name'] = 'Transformer'
if dataset == 'MIMIC':
c['no_diag'] = True
c['dataset'] = dataset
c = best_global(c)
# hyper-parameter grid
param_grid = {
'n_layers': [1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
'learning_rate':
list(np.logspace(np.log10(0.0001), np.log10(0.01), base=10, num=100)),
'batch_size':
list(
int(x)
for x in np.logspace(np.log2(4), np.log2(512), base=2, num=8)),
'trans_dropout_rate':
[0, 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5],
'd_model':
list(
int(x)
for x in np.logspace(np.log2(16), np.log2(256), base=2, num=5)),
'feedforward_size':
list(
int(x)
for x in np.logspace(np.log2(16), np.log2(256), base=2, num=5)),
'n_heads': [1, 2, 4, 8, 16]
}
c['n_layers'] = random.choice(param_grid['n_layers'])
c['learning_rate'] = round(random.choice(param_grid['learning_rate']), 5)
c['batch_size'] = random.choice(param_grid['batch_size'])
c['trans_dropout_rate'] = random.choice(param_grid['trans_dropout_rate'])
c['d_model'] = random.choice(param_grid['d_model'])
c['feedforward_size'] = random.choice(param_grid['feedforward_size'])
c['n_heads'] = random.choice(param_grid['n_heads'])
return c
from eICU_preprocessing.split_train_test import create_folder
from models.run_transformer import BaselineTransformer
from models.initialise_arguments import initialise_transformer_arguments
if __name__ == '__main__':
c = initialise_transformer_arguments()
c['mode'] = 'test'
c['exp_name'] = 'Transformer'
log_folder_path = create_folder('models/experiments/final', c.exp_name)
transformer = BaselineTransformer(
config=c,
n_epochs=c.n_epochs,
name=c.exp_name,
base_dir=log_folder_path,
explogger_kwargs={'folder_format': '%Y-%m-%d_%H%M%S{run_number}'})
transformer.run()
