WaveGRU-Text-To-Speech / extract_model.py
NTT123
first use of ema model.
17aa359
raw
history blame
186 Bytes
import pickle
dic = pickle.load(open("./wavegru_vocoder_1024_v3_1340000.ckpt", "rb"))
del dic["optim_state_dict"]
pickle.dump(dic, open("./wavegru_vocoder_1024_v3_1340000.ckpt", "wb"))