prepare_wsjmix,
kwargs={
"datapath": hparams["data_folder"],
"savepath": hparams["save_folder"],
"n_spks": hparams["num_spks"],
"skip_prep": hparams["skip_prep"],
},
)
# Create dataset objects
if hparams["dynamic_mixing"]:
if hparams["num_spks"] == 2:
from dynamic_mixing import dynamic_mix_data_prep # noqa
train_data = dynamic_mix_data_prep(hparams)
elif hparams["num_spks"] == 3:
from dynamic_mixing import dynamic_mix_data_prep_3mix # noqa
train_data = dynamic_mix_data_prep_3mix(hparams)
else:
raise ValueError(
"The specified number of speakers is not supported.")
_, valid_data, test_data = dataio_prep(hparams)
else:
train_data, valid_data, test_data = dataio_prep(hparams)
# Brain class initialization
separator = Separation(
modules=hparams["modules"],
opt_class=hparams["optimizer"],
)
# adjust the base_folder_dm path
hparams["base_folder_dm"] = (
os.path.normpath(hparams["base_folder_dm"]) + "_processed")
else:
print(
"Using the existing processed folder on the same directory as base_folder_dm"
)
hparams["base_folder_dm"] = (
os.path.normpath(hparams["base_folder_dm"]) + "_processed")
train_data = dynamic_mix_data_prep(
tr_csv=hparams["train_data"],
data_root_folder=hparams["data_folder"],
base_folder_dm=hparams["base_folder_dm"],
sample_rate=hparams["sample_rate"],
num_spks=hparams["num_spks"],
max_training_signal_len=hparams["training_signal_len"],
batch_size=hparams["dataloader_opts"]["batch_size"],
num_workers=hparams["dataloader_opts"]["num_workers"],
)
_, valid_data, test_data = dataio_prep(hparams)
else:
train_data, valid_data, test_data = dataio_prep(hparams)
# Load pretrained model if pretrained_separator is present in the yaml
if "pretrained_separator" in hparams:
run_on_main(hparams["pretrained_separator"].collect_files)
hparams["pretrained_separator"].load_collected()
# Brain class initialization