manifest_filepath_list=[args.valid_manifest_list[i]],
normalize=True,
augment=False,
input_type=args.input_type)
valid_sampler = BucketingSampler(valid_data, batch_size=args.k_train)
valid_loader = AudioDataLoader(pad_token_id=vocab.PAD_ID,
dataset=valid_data,
num_workers=args.num_workers)
valid_loader_list.append(valid_loader)
start_epoch = 0
metrics = None
loaded_args = None
logging.info("Continue from checkpoint:" + args.continue_from)
if args.training_mode == "meta":
model, vocab, _, _, epoch, metrics, loaded_args = load_meta_model(
args.continue_from)
else:
model, vocab, _, epoch, metrics, loaded_args = load_joint_model(
args.continue_from)
verbose = args.verbose
loss_type = args.loss
if USE_CUDA:
model = model.cuda()
logging.info(model)
num_epochs = args.epochs
print("Parameters: {}(trainable), {}(non-trainable)".format(
compute_num_params(model)[0],
manifest_filepath_list=[args.valid_manifest_list[i]],
normalize=True,
augment=False,
input_type=args.input_type)
valid_sampler = BucketingSampler(valid_data, batch_size=args.k_train)
valid_loader = AudioDataLoader(pad_token_id=vocab.PAD_ID,
dataset=valid_data,
num_workers=args.num_workers)
valid_loader_list.append(valid_loader)
start_epoch = 0
metrics = None
loaded_args = None
if args.continue_from != "":
logging.info("Continue from checkpoint:" + args.continue_from)
model, vocab, inner_opt, outer_opt, epoch, metrics, loaded_args = load_meta_model(
args.continue_from)
start_epoch = (epoch) # index starts from zero
verbose = args.verbose
else:
inner_opt, outer_opt = None, None
if args.model == "TRFS":
model = init_transformer_model(args,
vocab,
is_factorized=args.is_factorized,
r=args.r)
else:
logging.info("The model is not supported, check args --h")
loss_type = args.loss
if USE_CUDA: