checkpoint_dirpath = args.save_dirpath
if checkpoint_dirpath == 'checkpoints/':
checkpoint_dirpath += '%s+%s/%s' % (config["model"]["encoder"], config["model"]["decoder"], start_time)
if args.save_model:
summary_writer = SummaryWriter(log_dir=checkpoint_dirpath)
checkpoint_manager = CheckpointManager(model, optimizer, checkpoint_dirpath, config=config)
sparse_metrics = SparseGTMetrics()
ndcg = NDCG()
# If loading from checkpoint, adjust start epoch and load parameters.
if args.load_pthpath == "":
start_epoch = 0
else:
start_epoch = 0
model_state_dict, optimizer_state_dict = load_checkpoint(args.load_pthpath)
if isinstance(model, nn.DataParallel):
model.module.load_state_dict(model_state_dict)
else:
model.load_state_dict(model_state_dict)
print("Loaded model from {}".format(args.load_pthpath))
# =============================================================================
# TRAINING LOOP
# =============================================================================
# Forever increasing counter to keep track of iterations (for tensorboard log).
global_iteration_step = start_epoch * iterations
###load ndcg label list
samplefile = open('data/visdial_1.0_train_dense_sample.json', 'r')
# Pass vocabulary to construct Embedding layer.
encoder = Encoder(config["model"], val_dataset.vocabulary)
decoder = Decoder(config["model"], val_dataset.vocabulary)
print("Encoder: {}".format(config["model"]["encoder"]))
print("Decoder: {}".format(config["model"]["decoder"]))
# Share word embedding between encoder and decoder.
decoder.word_embed = encoder.word_embed
# Wrap encoder and decoder in a model.
model = EncoderDecoderModel(encoder, decoder).to(device)
if -1 not in args.gpu_ids:
model = nn.DataParallel(model, args.gpu_ids)
model_state_dict, _ = load_checkpoint(args.load_pthpath)
if isinstance(model, nn.DataParallel):
model.module.load_state_dict(model_state_dict)
else:
model.load_state_dict(model_state_dict)
print("Loaded model from {}".format(args.load_pthpath))
# Declare metric accumulators (won't be used if --split=test)
sparse_metrics = SparseGTMetrics()
ndcg = NDCG()
# =============================================================================
# EVALUATION LOOP
# =============================================================================
model.eval()