MAX_OUTPUT_LENGTH = 20
pd.set_option('display.width', 10000)
parser = argparse.ArgumentParser(description='')
parser.add_argument('--gpu', type=int, default=None)
parser.add_argument('--dataset_path', type=str)
parser.add_argument('--model_path', type=str)
args = parser.parse_args()
model_path = '../result/models/epoch3/model.ckpt'
dataset_path = '../data/dataset/train.csv' #################################### test.csv
dictionary_path = '../data/dataset/dictionary.pkl'
token2id = dataset.load_dictionary(dictionary_path)
id2token = {i:t for t, i in token2id.items()}
symbol_ids = {'<S>': token2id['<S>'], '<EOS>': token2id['<EOS>']}
vocab_size = len(list(token2id.keys()))
config.params.vocab_size = vocab_size
dataset = dataset.str2list(dataset.load_dataset(dataset_path, 1, 100))
sess = tf.Session()
if args.gpu:
with tf.device('/gpu:%d'%args.gpu):
model = ABSmodel(config.params)
model.rebuild_forward_graph(sess, model_path)
else:
model = ABSmodel(config.params)
parser.add_argument('--test_phase', type=bool, default=False)
args = parser.parse_args()
return args
if __name__ == '__main__':
args = parse_args()
torch.backends.cudnn.enabled = False
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
torch.backends.cudnn.benchmark = True
print('parameters:', args)
print('task:', args.task, 'model:', args.model)
# dictionary = Dictionary.load_from_file('./dictionary.pkl')
dictionary = load_dictionary(args.sentense_file_path, args.task)
train_dset = VQAFeatureDataset(args,
dictionary,
args.sentense_file_path,
args.feat_category,
args.feat_path,
mode='Train')
# val_dset = VQAFeatureDataset(args, dictionary, args.sentense_file_path,args.feat_category,args.feat_path, mode='Valid')
eval_dset = VQAFeatureDataset(args,
dictionary,
args.sentense_file_path,
args.feat_category,
args.feat_path,
mode='Test')
batch_size = args.batch_size
import dataset from nets import simpleNet from ma import movingAverage from sklearn.metrics import classification_report y_true = [] y_pred = [] def monitorMA(example, movingAverage): if example not in monitor: monitor[example] = list() else: monitor[example].append(movingAverage) data = dataset.load_dictionary() nn = simpleNet(architecture=numpy.array([15 , data.shape[0]])) success_rate = dict() monitor = dict() for epoch in tqdm(range(50000)): for example in range(data.shape[0]): input = data[example][1] target = data[example][0] y_true.append(target) output = nn.forward(input) y_pred.append(output)