deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-05 18:09:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix bug

Browse Source
This commit is contained in:
lyuxiang.lx
2024-11-22 15:11:55 +08:00
parent c3dfd23399
commit dc3f6432ba
2 changed files with 7 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
cosyvoice/cli/cosyvoice.py
View File

@@ -16,6 +16,7 @@ import time
from tqdm import tqdm from tqdm import tqdm
from hyperpyyaml import load_hyperpyyaml from hyperpyyaml import load_hyperpyyaml
from modelscope import snapshot_download from modelscope import snapshot_download
import torch
from cosyvoice.cli.frontend import CosyVoiceFrontEnd from cosyvoice.cli.frontend import CosyVoiceFrontEnd
from cosyvoice.cli.model import CosyVoiceModel from cosyvoice.cli.model import CosyVoiceModel
from cosyvoice.utils.file_utils import logging from cosyvoice.utils.file_utils import logging
@@ -37,6 +38,10 @@ class CosyVoice:
'{}/spk2info.pt'.format(model_dir), '{}/spk2info.pt'.format(model_dir),
instruct, instruct,
configs['allowed_special']) configs['allowed_special'])
if torch.cuda.is_available() is False and (fp16 is True or load_jit is True):
load_jit = False
fp16 = False
logging.warning('cpu do not support fp16 and jit, force set to False')
self.model = CosyVoiceModel(configs['llm'], configs['flow'], configs['hift'], fp16) self.model = CosyVoiceModel(configs['llm'], configs['flow'], configs['hift'], fp16)
self.model.load('{}/llm.pt'.format(model_dir), self.model.load('{}/llm.pt'.format(model_dir),
'{}/flow.pt'.format(model_dir), '{}/flow.pt'.format(model_dir),

2
cosyvoice/utils/common.py
View File

@@ -141,6 +141,8 @@ def fade_in_out(fade_in_mel, fade_out_mel, window):
device = fade_in_mel.device device = fade_in_mel.device
fade_in_mel, fade_out_mel = fade_in_mel.cpu(), fade_out_mel.cpu() fade_in_mel, fade_out_mel = fade_in_mel.cpu(), fade_out_mel.cpu()
mel_overlap_len = int(window.shape[0] / 2) mel_overlap_len = int(window.shape[0] / 2)
if fade_in_mel.device == torch.device('cpu'):
fade_in_mel = fade_in_mel.clone()
fade_in_mel[..., :mel_overlap_len] = fade_in_mel[..., :mel_overlap_len] * window[:mel_overlap_len] + \ fade_in_mel[..., :mel_overlap_len] = fade_in_mel[..., :mel_overlap_len] * window[:mel_overlap_len] + \
fade_out_mel[..., -mel_overlap_len:] * window[mel_overlap_len:] fade_out_mel[..., -mel_overlap_len:] * window[mel_overlap_len:]
return fade_in_mel.to(device) return fade_in_mel.to(device)