What version of torch？

[aijianiula0601]

What version of the following module is? Please give us the detailed version.

torch
torchvision
torchaudio
torchtext

I have the following problem：
/python3.11/site-packages/torchtext/lib/libtorchtext.so: undefined symbol: _ZN5torch3jit21setUTF8DecodingIgnoreEb

And what's your cuda version? Thank you

[GiJeongCho]

me too.. i saw same erorr
my versions
conda install torchtext==0.17.0
pip install torch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 --index-url https://download.pytorch.org/whl/cu121

OSError: /home/edutem/anaconda3/envs/vits2/lib/python3.11/site-packages/torchtext/lib/libtorchtext.so: undefined symbol: _ZN3c109TupleTypeC1ESt6vectorINS_4Type24SingletonOrSharedTypePtrIS2_EESaIS4_EESt8optionalINS_13QualifiedNameEESt10shared_ptrINS_14FunctionSchemaEE

[Wyzix33]

i was able to run it using pip install torch==2.1.0+cu121 torchvision==0.16.0+cu121 torchaudio==2.1.0+cu121 torchtext==0.16.0 -f https://download.pytorch.org/whl/torch_stable.html
and pip install numpy==1.26.4
also the interface was not working, i ended up doing this:

import IPython.display as ipd
import torch
from torch.utils.data import DataLoader
from text import tokenizer
from torchtext.vocab import Vocab

from utils.task import load_vocab
from text.symbols import  UNK_ID
from utils.task import load_checkpoint
from utils.hparams import get_hparams_from_file
from data_utils import TextAudioSpeakerLoader, TextAudioSpeakerCollate
from model.models import SynthesizerTrn
from text.symbols import symbols
from text import tokenizer



model = "romanian_base"
checkpoint = "G_87000.pth"
hps = get_hparams_from_file(f"./datasets/{model}/config.yaml")
text_cleaners = hps.data.text_cleaners
vocab_file = f"./datasets/{model}/vocab.txt"
token_idx = text_cleaners.index("tokenize_text")
token_cleaners = text_cleaners[token_idx:]

def separate_text_cleaners(text_cleaners):
    final_list = []
    temp_list = []

    for cleaner in text_cleaners:
        if cleaner == "phonemize_text":
            if temp_list:
                final_list.append(temp_list)
            final_list.append([cleaner])
            temp_list = []
        else:
            temp_list.append(cleaner)

    if temp_list:
        final_list.append(temp_list)

    return final_list


text_cleaners = text_cleaners[:token_idx]
text_cleaners = separate_text_cleaners(text_cleaners)
vocab = load_vocab(vocab_file)

filter_length = hps.data.n_mels if hps.data.use_mel else hps.data.n_fft // 2 + 1
segment_size = hps.train.segment_size // hps.data.hop_length
net_g = SynthesizerTrn(len(symbols), filter_length, segment_size, **hps.model).cuda()
_ = net_g.eval()
_ = load_checkpoint(f"./datasets/{model}/logs/{checkpoint}", net_g, None)

text = "example text for testing"
for cleaners in text_cleaners:
    text = tokenizer(text, Vocab, cleaners, language=hps.data.language)
temp = tokenizer(text, vocab, token_cleaners, language=hps.data.language)
assert UNK_ID not in temp, f"Found unknown symbol:\n{text}\n{detokenizer(temp)}"
print(temp)
stn_tst =  torch.LongTensor(temp)
with torch.no_grad():
    x_tst = stn_tst.cuda().unsqueeze(0)
    x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()

    out = net_g.infer(x_tst, x_tst_lengths, noise_scale=0.667, noise_scale_w=0.8, length_scale=1)
    audio = out[0][0, 0].data.cpu().float().numpy()
ipd.display(ipd.Audio(audio, rate=hps.data.sample_rate, normalize=False))

[annam-airia]

i was able to run it using pip install torch==2.1.0+cu121 torchvision==0.16.0+cu121 torchaudio==2.1.0+cu121 torchtext==0.16.0 -f https://download.pytorch.org/whl/torch_stable.html and pip install numpy==1.26.4 also the interface was not working, i ended up doing this:

import IPython.display as ipd
import torch
from torch.utils.data import DataLoader
from text import tokenizer
from torchtext.vocab import Vocab

from utils.task import load_vocab
from text.symbols import  UNK_ID
from utils.task import load_checkpoint
from utils.hparams import get_hparams_from_file
from data_utils import TextAudioSpeakerLoader, TextAudioSpeakerCollate
from model.models import SynthesizerTrn
from text.symbols import symbols
from text import tokenizer



model = "romanian_base"
checkpoint = "G_87000.pth"
hps = get_hparams_from_file(f"./datasets/{model}/config.yaml")
text_cleaners = hps.data.text_cleaners
vocab_file = f"./datasets/{model}/vocab.txt"
token_idx = text_cleaners.index("tokenize_text")
token_cleaners = text_cleaners[token_idx:]

def separate_text_cleaners(text_cleaners):
    final_list = []
    temp_list = []

    for cleaner in text_cleaners:
        if cleaner == "phonemize_text":
            if temp_list:
                final_list.append(temp_list)
            final_list.append([cleaner])
            temp_list = []
        else:
            temp_list.append(cleaner)

    if temp_list:
        final_list.append(temp_list)

    return final_list


text_cleaners = text_cleaners[:token_idx]
text_cleaners = separate_text_cleaners(text_cleaners)
vocab = load_vocab(vocab_file)

filter_length = hps.data.n_mels if hps.data.use_mel else hps.data.n_fft // 2 + 1
segment_size = hps.train.segment_size // hps.data.hop_length
net_g = SynthesizerTrn(len(symbols), filter_length, segment_size, **hps.model).cuda()
_ = net_g.eval()
_ = load_checkpoint(f"./datasets/{model}/logs/{checkpoint}", net_g, None)

text = "example text for testing"
for cleaners in text_cleaners:
    text = tokenizer(text, Vocab, cleaners, language=hps.data.language)
temp = tokenizer(text, vocab, token_cleaners, language=hps.data.language)
assert UNK_ID not in temp, f"Found unknown symbol:\n{text}\n{detokenizer(temp)}"
print(temp)
stn_tst =  torch.LongTensor(temp)
with torch.no_grad():
    x_tst = stn_tst.cuda().unsqueeze(0)
    x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()

    out = net_g.infer(x_tst, x_tst_lengths, noise_scale=0.667, noise_scale_w=0.8, length_scale=1)
    audio = out[0][0, 0].data.cpu().float().numpy()
ipd.display(ipd.Audio(audio, rate=hps.data.sample_rate, normalize=False))

Hi Wyzix33, thanks for the help. Wondering how you delt with the import "from text.symbols import symbols" when the content for symbols.py is:

"""
Set of symbols
"""
_punctuation = ';:,.!?¡¿—…"«»“”'

"""
Special symbols
"""

special_symbols = ["", "", "", "", "", ""]
PAD_ID, UNK_ID, BOS_ID, EOS_ID = 0, 1, 2, 3

... without 'symbols'.