5
Run: conda install -c conda-forge pynini=2.1.5 && pip install WeTextProcessing
Traceback (most recent call last):
File "/home/usr/Documents/git/ChatTTS/ChatTTS.py", line 9, in
Q
[2noise/ChatTTS]UnboundLocalError：无法访问与值无关的局部变量“Normalizer”
A
回答
9
pip install WeTextProcessing
3
Collecting WeTextProcessing
Using cached WeTextProcessing-0.1.12-py3-none-any.whl.metadata (5.2 kB)
Collecting pynini==2.1.5 (from WeTextProcessing)
Using cached pynini-2.1.5.tar.gz (627 kB)
Preparing metadata (setup.py) ... error
error: subprocess-exited-with-error
× python setup.py egg_info did not run successfully.
│ exit code: 1
╰─> [6 lines of output]
Traceback (most recent call last):
File "
note: This error originates from a subprocess, and is likely not a problem with pip. error: metadata-generation-failed
× Encountered error while generating package metadata.
9
(.venv) root@This-is-Python ChatTTS % pip install WeTextProcessing Looking in indexes: https://mirrors.aliyun.com/pypi/simple/ Collecting WeTextProcessing Using cached https://mirrors.aliyun.com/pypi/packages/04/17/3d624a59ef30484c0f4bb703529f128566fab0e8660bbd2e3a36f5244d2d/WeTextProcessing-0.1.12-py3-none-any.whl (690 kB) Collecting pynini==2.1.5 (from WeTextProcessing) Using cached https://mirrors.aliyun.com/pypi/packages/86/63/6a720dbdf4e7358baa2ab206a51693f9c6e83d174179a95539f501bd4a34/pynini-2.1.5.tar.gz (627 kB) Preparing metadata (setup.py) ... done Requirement already satisfied: importlib-resources in ./.venv/lib/python3.11/site-packages (from WeTextProcessing) (6.4.0) Requirement already satisfied: Cython>=0.29 in ./.venv/lib/python3.11/site-packages (from pynini==2.1.5->WeTextProcessing) (3.0.10) Building wheels for collected packages: pynini Building wheel for pynini (setup.py) ... error error: subprocess-exited-with-error
× python setup.py bdist_wheel did not run successfully. │ exit code: 1 ╰─> [55 lines of output] running bdist_wheel running build running build_py creating build creating build/lib.macosx-13.0-arm64-cpython-311 creating build/lib.macosx-13.0-arm64-cpython-311/pywrapfst copying pywrapfst/init.py -> build/lib.macosx-13.0-arm64-cpython-311/pywrapfst creating build/lib.macosx-13.0-arm64-cpython-311/pynini copying pynini/init.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini creating build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/examples/chatspeak.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/examples/chatspeak_model.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/examples/weather.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/examples/init.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/examples/numbers.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/examples/plurals.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/examples/t9.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/examples/case.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/examples/dates.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/examples/g2p.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples creating build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/lib/edit_transducer.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/lib/utf8.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/lib/init.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/lib/features.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/lib/pynutil.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/lib/paradigms.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/lib/tagger.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/lib/rule_cascade.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/lib/rewrite.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/lib/byte.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib creating build/lib.macosx-13.0-arm64-cpython-311/pynini/export copying pynini/export/init.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/export copying pynini/export/export.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/export copying pynini/export/multi_grm_example.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/export copying pynini/export/multi_grm.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/export copying pynini/export/grm.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/export copying pynini/export/grm_example.py -> build/lib.macosx-13.0-arm64-cpython-311/pynini/export copying pywrapfst/init.pyi -> build/lib.macosx-13.0-arm64-cpython-311/pywrapfst copying pywrapfst/py.typed -> build/lib.macosx-13.0-arm64-cpython-311/pywrapfst copying pynini/init.pyi -> build/lib.macosx-13.0-arm64-cpython-311/pynini copying pynini/py.typed -> build/lib.macosx-13.0-arm64-cpython-311/pynini copying pynini/examples/py.typed -> build/lib.macosx-13.0-arm64-cpython-311/pynini/examples copying pynini/lib/py.typed -> build/lib.macosx-13.0-arm64-cpython-311/pynini/lib copying pynini/export/py.typed -> build/lib.macosx-13.0-arm64-cpython-311/pynini/export running build_ext building '_pywrapfst' extension creating build/temp.macosx-13.0-arm64-cpython-311 creating build/temp.macosx-13.0-arm64-cpython-311/extensions clang -Wsign-compare -Wunreachable-code -fno-common -dynamic -DNDEBUG -g -fwrapv -O3 -Wall -isysroot /Library/Developer/CommandLineTools/SDKs/MacOSX13.sdk -I/Users/guoge/Work/DEV/server/chattts/ChatTTS/.venv/include -I/opt/homebrew/opt/python@3.11/Frameworks/Python.framework/Versions/3.11/include/python3.11 -c extensions/_pywrapfst.cpp -o build/temp.macosx-13.0-arm64-cpython-311/extensions/_pywrapfst.o -std=c++17 -Wno-register -Wno-deprecated-declarations -Wno-unused-function -Wno-unused-local-typedefs -funsigned-char -stdlib=libc++ -mmacosx-version-min=10.7 extensions/_pywrapfst.cpp:1291:10: fatal error: 'fst/util.h' file not found
include <fst/util.h> ^~~~~~~~~~~~
1 error generated.
error: command '/usr/bin/clang' failed with exit code 1
[end of output]note: This error originates from a subprocess, and is likely not a problem with pip. ERROR: Failed building wheel for pynini Running setup.py clean for pynini Failed to build pynini ERROR: Could not build wheels for pynini, which is required to install pyproject.toml-based projects
环境： python@3.11 pip 24.0 MAC OS 装不上pynini，WeTextProcessing也没有办法加载。
3
用miniconda跑吧
2
Traceback (most recent call last):
File "/home/brandon/Documents/git/ChatTTS/ChatTTS.py", line 11, in
6
Traceback (most recent call last):
File "/home/brandon/Documents/git/ChatTTS/ChatTTS.py", line 11, in
7
Core.py 修改后，可以run起来。注释掉就可以。 `import os import logging from functools import partial from omegaconf import OmegaConf
import torch from vocos import Vocos from .model.dvae import DVAE from .model.gpt import GPT_warpper from .utils.gpu_utils import select_device from .utils.infer_utils import count_invalid_characters, detect_language, apply_character_map, apply_half2full_map from .utils.io_utils import get_latest_modified_file from .infer.api import refine_text, infer_code
from huggingface_hub import snapshot_download
logging.basicConfig(level = logging.INFO)
class Chat: def init(self, ): self.pretrain_models = {} self.normalizer = {} self.logger = logging.getLogger(name)
def check_model(self, level = logging.INFO, use_decoder = False):
not_finish = False
check_list = ['vocos', 'gpt', 'tokenizer']
if use_decoder:
check_list.append('decoder')
else:
check_list.append('dvae')
for module in check_list:
if module not in self.pretrain_models:
self.logger.log(logging.WARNING, f'{module} not initialized.')
not_finish = True
if not not_finish:
self.logger.log(level, f'All initialized.')
return not not_finish
def load_models(self, source='huggingface', force_redownload=False, local_path='<LOCAL_PATH>', **kwargs):
if source == 'huggingface':
hf_home = os.getenv('HF_HOME', os.path.expanduser("~/.cache/huggingface"))
try:
download_path = get_latest_modified_file(os.path.join(hf_home, 'hub/models--2Noise--ChatTTS/snapshots'))
except:
download_path = None
if download_path is None or force_redownload:
self.logger.log(logging.INFO, f'Download from HF: https://huggingface.co/2Noise/ChatTTS')
download_path = snapshot_download(repo_id="2Noise/ChatTTS", allow_patterns=["*.pt", "*.yaml"])
else:
self.logger.log(logging.INFO, f'Load from cache: {download_path}')
elif source == 'local':
self.logger.log(logging.INFO, f'Load from local: {local_path}')
download_path = local_path
self._load(**{k: os.path.join(download_path, v) for k, v in OmegaConf.load(os.path.join(download_path, 'config', 'path.yaml')).items()}, **kwargs)
def _load(
self,
vocos_config_path: str = None,
vocos_ckpt_path: str = None,
dvae_config_path: str = None,
dvae_ckpt_path: str = None,
gpt_config_path: str = None,
gpt_ckpt_path: str = None,
decoder_config_path: str = None,
decoder_ckpt_path: str = None,
tokenizer_path: str = None,
device: str = None,
compile: bool = True,
):
if not device:
device = select_device(4096)
self.logger.log(logging.INFO, f'use {device}')
if vocos_config_path:
vocos = Vocos.from_hparams(vocos_config_path).to(device).eval()
assert vocos_ckpt_path, 'vocos_ckpt_path should not be None'
vocos.load_state_dict(torch.load(vocos_ckpt_path))
self.pretrain_models['vocos'] = vocos
self.logger.log(logging.INFO, 'vocos loaded.')
if dvae_config_path:
cfg = OmegaConf.load(dvae_config_path)
dvae = DVAE(**cfg).to(device).eval()
assert dvae_ckpt_path, 'dvae_ckpt_path should not be None'
dvae.load_state_dict(torch.load(dvae_ckpt_path, map_location='cpu'))
self.pretrain_models['dvae'] = dvae
self.logger.log(logging.INFO, 'dvae loaded.')
if gpt_config_path:
cfg = OmegaConf.load(gpt_config_path)
gpt = GPT_warpper(**cfg).to(device).eval()
assert gpt_ckpt_path, 'gpt_ckpt_path should not be None'
gpt.load_state_dict(torch.load(gpt_ckpt_path, map_location='cpu'))
if compile and 'cuda' in str(device):
gpt.gpt.forward = torch.compile(gpt.gpt.forward, backend='inductor', dynamic=True)
self.pretrain_models['gpt'] = gpt
spk_stat_path = os.path.join(os.path.dirname(gpt_ckpt_path), 'spk_stat.pt')
assert os.path.exists(spk_stat_path), f'Missing spk_stat.pt: {spk_stat_path}'
self.pretrain_models['spk_stat'] = torch.load(spk_stat_path).to(device)
self.logger.log(logging.INFO, 'gpt loaded.')
if decoder_config_path:
cfg = OmegaConf.load(decoder_config_path)
decoder = DVAE(**cfg).to(device).eval()
assert decoder_ckpt_path, 'decoder_ckpt_path should not be None'
decoder.load_state_dict(torch.load(decoder_ckpt_path, map_location='cpu'))
self.pretrain_models['decoder'] = decoder
self.logger.log(logging.INFO, 'decoder loaded.')
if tokenizer_path:
tokenizer = torch.load(tokenizer_path, map_location='cpu')
tokenizer.padding_side = 'left'
self.pretrain_models['tokenizer'] = tokenizer
self.logger.log(logging.INFO, 'tokenizer loaded.')
self.check_model()
def infer(
self,
text,
skip_refine_text=False,
refine_text_only=False,
params_refine_text={},
params_infer_code={'prompt':'[speed_5]'},
use_decoder=True,
do_text_normalization=True,
lang=None,
):
assert self.check_model(use_decoder=use_decoder)
if not isinstance(text, list):
text = [text]
# if do_text_normalization:
# for i, t in enumerate(text):
# _lang = detect_language(t) if lang is None else lang
# self.init_normalizer(_lang)
# text[i] = self.normalizer[_lang](t)
# if _lang == 'zh':
# text[i] = apply_half2full_map(text[i])
for i, t in enumerate(text):
invalid_characters = count_invalid_characters(t)
if len(invalid_characters):
self.logger.log(logging.WARNING, f'Invalid characters found! : {invalid_characters}')
text[i] = apply_character_map(t)
if not skip_refine_text:
text_tokens = refine_text(self.pretrain_models, text, **params_refine_text)['ids']
text_tokens = [i[i < self.pretrain_models['tokenizer'].convert_tokens_to_ids('[break_0]')] for i in text_tokens]
text = self.pretrain_models['tokenizer'].batch_decode(text_tokens)
if refine_text_only:
return text
text = [params_infer_code.get('prompt', '') + i for i in text]
params_infer_code.pop('prompt', '')
result = infer_code(self.pretrain_models, text, **params_infer_code, return_hidden=use_decoder)
if use_decoder:
mel_spec = [self.pretrain_models['decoder'](i[None].permute(0,2,1)) for i in result['hiddens']]
else:
mel_spec = [self.pretrain_models['dvae'](i[None].permute(0,2,1)) for i in result['ids']]
wav = [self.pretrain_models['vocos'].decode(i).cpu().numpy() for i in mel_spec]
return wav
def sample_random_speaker(self, ):
dim = self.pretrain_models['gpt'].gpt.layers[0].mlp.gate_proj.in_features
std, mean = self.pretrain_models['spk_stat'].chunk(2)
return torch.randn(dim, device=std.device) * std + mean
# def init_normalizer(self, lang):
# if lang not in self.normalizer:
# if lang == 'zh':
# try:
# from tn.chinese.normalizer import Normalizer
# except:
# self.logger.log(logging.WARNING, f'Package WeTextProcessing not found! \
# Run: conda install -c conda-forge pynini=2.1.5 && pip install WeTextProcessing')
# self.normalizer[lang] = Normalizer().normalize
# else:
# try:
# from nemo_text_processing.text_normalization.normalize import Normalizer
# except:
# self.logger.log(logging.WARNING, f'Package nemo_text_processing not found! \
# Run: conda install -c conda-forge pynini=2.1.5 && pip install nemo_text_processing')
# self.normalizer[lang] = partial(Normalizer(input_case='cased', lang=lang).normalize, verbose=False, punct_post_process=True)`
9
注释掉两处还是之前的问题，没用。
5
我复制它的示例代码
import ChatTTS
from IPython.display import Audio
chat = ChatTTS.Chat()
chat.load_models(compile=False) # 设置为True以获得更快速度
texts = ["在这里输入你的文本",]
wavs = chat.infer(texts, use_decoder=True)
torchaudio.save("output1.wav", torch.from_numpy(wavs[0]), 24000)应该是可以直接运行的吗？还是要修改一下？
5
我复制它的示例代码
import ChatTTS from IPython.display import Audio chat = ChatTTS.Chat() chat.load_models(compile=False) # 设置为True以获得更快速度 texts = ["在这里输入你的文本",] wavs = chat.infer(texts, use_decoder=True) torchaudio.save("output1.wav", torch.from_numpy(wavs[0]), 24000)应该是可以直接运行的吗？还是要修改一下？
你直接download官方的整个项目可以正常run。 ！注销这两处，就关闭了english -> chinese的功能，在文本中的数字、英文无法转译成中文语音朗读。会出现许多[cat]。
下载整个项目，选择python 3.11运行，安装.ven环境依赖（不要用全局） 然后执行python webui.py
9
也可以直接关闭Core.py中的开关do_text_normalization，置为False：
def infer(
self,
text,
skip_refine_text=False,
refine_text_only=False,
params_refine_text={},
params_infer_code={'prompt':'[speed_5]'},
use_decoder=True,
do_text_normalization=False,
lang=None,
):