WaveGRU-Text-To-Speech / extract_model.py
NTT123
use a customized gru.
41ba53f
raw
history blame
138 Bytes
import pickle
dic = pickle.load(open("./wavegru.ckpt", "rb"))
del dic["optim_state_dict"]
pickle.dump(dic, open("./wavegru.ckpt", "wb"))