Open SeoheeJang1 opened 1 year ago
Dear SCIANN team
when I use model.train, I use adaptive weight to train, but there is another option which is called as target weight
if I use both, then can these be used simultaneously, or one of them override somehow??
like this
H_model = model.train(input_data, model_target_data, target_weights=[1,1,1,1,10,10,100,100] learning_rate=([0, 10000,20000,30000,40000,50000], [0.001, 0.0005, 0.0001, 0.0001, 0.0001, 0.0001]), batch_size=BATCH_SIZE, epochs=EPOCHS_model, stop_loss_value=1e-8, stop_after=STOP_AFTER, verbose=1, adaptive_weights={'method': 'ntk', 'freq':5000}, )
Best Regards
Seohee Jang
Dear SCIANN team
when I use model.train, I use adaptive weight to train, but there is another option which is called as target weight
if I use both, then can these be used simultaneously, or one of them override somehow??
like this
H_model = model.train(input_data, model_target_data, target_weights=[1,1,1,1,10,10,100,100] learning_rate=([0, 10000,20000,30000,40000,50000], [0.001, 0.0005, 0.0001, 0.0001, 0.0001, 0.0001]), batch_size=BATCH_SIZE, epochs=EPOCHS_model, stop_loss_value=1e-8, stop_after=STOP_AFTER, verbose=1, adaptive_weights={'method': 'ntk', 'freq':5000}, )
Best Regards
Seohee Jang