assert cfg.EXP_NAME == os.path.basename(args.cfg).replace('.yaml', '')
if args.opts:
merge_cfg_from_list(args.opts)
# Start session
os.environ["CUDA_VISIBLE_DEVICES"] = str(cfg.GPU_ID)
sess = tf.Session(config=tf.ConfigProto(
gpu_options=tf.GPUOptions(allow_growth=cfg.GPU_MEM_GROWTH)))
# Data files
imdb_file = cfg.IMDB_FILE % cfg.TRAIN.SPLIT_VQA
data_reader = DataReader(
imdb_file, shuffle=True, one_pass=False, batch_size=cfg.TRAIN.BATCH_SIZE,
vocab_question_file=cfg.VOCAB_QUESTION_FILE, T_encoder=cfg.MODEL.T_ENCODER,
vocab_answer_file=cfg.VOCAB_ANSWER_FILE,
load_gt_layout=cfg.TRAIN.USE_GT_LAYOUT,
vocab_layout_file=cfg.VOCAB_LAYOUT_FILE, T_decoder=cfg.MODEL.T_CTRL,
load_soft_score=cfg.TRAIN.VQA_USE_SOFT_SCORE)
num_vocab = data_reader.batch_loader.vocab_dict.num_vocab
num_choices = data_reader.batch_loader.answer_dict.num_vocab
module_names = data_reader.batch_loader.layout_dict.word_list
# Inputs and model
input_seq_batch = tf.placeholder(tf.int32, [None, None])
seq_length_batch = tf.placeholder(tf.int32, [None])
image_feat_batch = tf.placeholder(
tf.float32, [None, cfg.MODEL.H_FEAT, cfg.MODEL.W_FEAT, cfg.MODEL.FEAT_DIM])
model = Model(
input_seq_batch, seq_length_batch, image_feat_batch, num_vocab=num_vocab,
num_choices=num_choices, module_names=module_names, is_training=True)
save_file = './exp_vqa/results/%s/%s.%s.txt' % (exp_name, snapshot_name,
tst_image_set)
os.makedirs(os.path.dirname(save_file), exist_ok=True)
eval_output_name = 'vqa_OpenEnded_mscoco_%s_%s_%s_results.json' % (
tst_image_set, exp_name, snapshot_name)
eval_output_file = './exp_vqa/eval_outputs/%s/%s' % (exp_name,
eval_output_name)
os.makedirs(os.path.dirname(eval_output_file), exist_ok=True)
assembler = Assembler(vocab_layout_file)
data_reader_tst = DataReader(imdb_file_tst,
shuffle=False,
one_pass=True,
batch_size=N,
T_encoder=T_encoder,
T_decoder=T_decoder,
assembler=assembler,
vocab_question_file=vocab_question_file,
vocab_answer_file=vocab_answer_file)
num_vocab_txt = data_reader_tst.batch_loader.vocab_dict.num_vocab
num_vocab_nmn = len(assembler.module_names)
num_choices = data_reader_tst.batch_loader.answer_dict.num_vocab
# Network inputs
input_seq_batch = tf.placeholder(tf.int32, [None, None])
seq_length_batch = tf.placeholder(tf.int32, [None])
image_feat_batch = tf.placeholder(tf.float32, [None, H_feat, W_feat, D_feat])
expr_validity_batch = tf.placeholder(tf.bool, [None])
assert cfg.EXP_NAME == os.path.basename(args.cfg).replace('.yaml', '')
if args.opts:
merge_cfg_from_list(args.opts)
# Start session
os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID" ### edit_vedika
os.environ["CUDA_VISIBLE_DEVICES"] = str(cfg.GPU_ID)
sess = tf.Session(config=tf.ConfigProto(
gpu_options=tf.GPUOptions(allow_growth=cfg.GPU_MEM_GROWTH)))
# Data files
imdb_file = cfg.IMDB_FILE % cfg.TEST.SPLIT_VQA
#print(imdb_file)
data_reader = DataReader(
imdb_file, shuffle=False, one_pass=True, batch_size=cfg.TRAIN.BATCH_SIZE, ## cfg.TRAIN.BATCH_SIZE =128 (vqa_v2_scratch.yaml)
vocab_question_file=cfg.VOCAB_QUESTION_FILE, T_encoder=cfg.MODEL.T_ENCODER,
vocab_answer_file=cfg.VOCAB_ANSWER_FILE, load_gt_layout=True,
vocab_layout_file=cfg.VOCAB_LAYOUT_FILE, T_decoder=cfg.MODEL.T_CTRL)
num_vocab = data_reader.batch_loader.vocab_dict.num_vocab
num_choices = data_reader.batch_loader.answer_dict.num_vocab
module_names = data_reader.batch_loader.layout_dict.word_list
# Eval files
if cfg.TEST.GEN_EVAL_FILE:
eval_file_pkl = cfg.TEST.EVAL_FILE_pkl % (
cfg.EXP_NAME, cfg.TEST.SPLIT_VQA, cfg.EXP_NAME, cfg.TEST.ITER)
# print('evaluation outputs will be saved to %s' % eval_file_pkl)
os.makedirs(os.path.dirname(eval_file_pkl), exist_ok=True)
answer_word_list = data_reader.batch_loader.answer_dict.word_list ##aanswer_word_list is here
assert(answer_word_list[0] == '<unk>') ### here asserting unk to position 0