'late_conv_dict': {
'conv_filter_list': [(512, 1), (512, 1)],
'pool_filter_list': [None, None],
'pool_stride_list': [None, None],
},
'dense_filter_size': 1,
'final_pool_function': T.mean,
'input_size': 128,
'output_size': n_top,
'p_dropout': 0.5
}
# Loading data
print("Loading data...")
X_tr, y_tr = utils.load_data('tr',
exp_data_dir=exp_data_dir,
use_real_data=use_real_data)
X_va, y_va = utils.load_data('va',
exp_data_dir=exp_data_dir,
use_real_data=use_real_data)
network, input_var, lr_var, train_func, val_func, pr_func = \
utils.make_network(
network_type, loss_function, lr, network_options
)
# Training
utils.train(X_tr,
y_tr,
X_va,
y_va,
},
'late_conv_dict': {
'conv_filter_list': [(512, 1), (512, 1)],
'pool_filter_list': [None, None],
'pool_stride_list': [None, None],
},
'dense_filter_size': 1,
'final_pool_function': T.mean,
'input_size': 128,
'output_size': 50,
'p_dropout': 0.5
}
# Load data
X_te, y_te = utils.load_data('te',
exp_data_dir=exp_data_dir,
use_real_data=True)
# Make network
network, input_var, pr_func = utils.make_network_test(
network_type, network_options)
fn_list = sorted(os.listdir(model_dir))
score_list = list()
for fn in fn_list:
print('Processing model {}'.format(fn))
param_fp = os.path.join(model_dir, fn)
# Load params
utils.load_model(param_fp, network)