Spaces:
Running
Running
import argparse | |
import random | |
import numpy as np | |
import torch | |
from trainer import CBHGTrainer, Seq2SeqTrainer, GPTTrainer | |
SEED = 1234 | |
random.seed(SEED) | |
np.random.seed(SEED) | |
torch.manual_seed(SEED) | |
torch.cuda.manual_seed(SEED) | |
torch.backends.cudnn.deterministic = True | |
torch.backends.cudnn.benchmark = False | |
def train_parser(): | |
parser = argparse.ArgumentParser() | |
parser.add_argument("--model_kind", dest="model_kind", type=str, required=True) | |
parser.add_argument( | |
"--model_desc", dest="model_desc", type=str, required=False, default="" | |
) | |
parser.add_argument("--config", dest="config", type=str, required=True) | |
parser.add_argument( | |
"--reset_dir", | |
dest="clear_dir", | |
action="store_true", | |
help="deletes everything under this config's folder.", | |
) | |
return parser | |
parser = train_parser() | |
args = parser.parse_args() | |
if args.model_kind in ["seq2seq"]: | |
trainer = Seq2SeqTrainer(args.config, args.model_kind, args.model_desc) | |
elif args.model_kind in ["tacotron_based"]: | |
trainer = Seq2SeqTrainer(args.config, args.model_kind, args.model_desc) | |
elif args.model_kind in ["baseline", "cbhg"]: | |
trainer = CBHGTrainer(args.config, args.model_kind, args.model_desc) | |
elif args.model_kind in ["gpt"]: | |
trainer = GPTTrainer(args.config, args.model_kind, args.model_desc) | |
else: | |
raise ValueError("The model kind is not supported") | |
trainer.run() | |