mrq/vall-e
5
9
Fork 0
Code Issues 8 Pull Requests Packages Projects Releases Wiki Activity

should not be working late

Browse Source
This commit is contained in:
mrq 2025-04-03 23:32:58 -05:00
parent bfe70e9d56
commit 28d39ef962
2 changed files with 7 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
vall_e/emb/qnt.py
View File

@ -20,32 +20,29 @@ from tqdm import tqdm
from torch.nn.utils.rnn import pad_sequence from torch.nn.utils.rnn import pad_sequence
AVAILABLE_AUDIO_BACKENDS = []
try: try:
from .codecs.encodec import * from .codecs.encodec import *
cfg.inference.audio_backends["encodec"] = True AVAILABLE_AUDIO_BACKENDS.append("encodec")
except Exception as e: except Exception as e:
cfg.inference.audio_backends["encodec"] = False # e
_logger.warning(str(e)) _logger.warning(str(e))
try: try:
from .codecs.vocos import * from .codecs.vocos import *
cfg.inference.audio_backends["vocos"] = True AVAILABLE_AUDIO_BACKENDS.append("vocos")
except Exception as e: except Exception as e:
cfg.inference.audio_backends["vocos"] = False # e
_logger.warning(str(e)) _logger.warning(str(e))
try: try:
from .codecs.dac import * from .codecs.dac import *
cfg.inference.audio_backends["dac"] = True AVAILABLE_AUDIO_BACKENDS.append("dac")
except Exception as e: except Exception as e:
cfg.inference.audio_backends["dac"] = False # e
_logger.warning(str(e)) _logger.warning(str(e))
try: try:
from .codecs.nemo import * from .codecs.nemo import *
cfg.inference.audio_backends["nemo"] = True AVAILABLE_AUDIO_BACKENDS.append("nemo")
except Exception as e: except Exception as e:
cfg.inference.audio_backends["nemo"] = False # e
_logger.warning(str(e)) _logger.warning(str(e))
@cache @cache

4
vall_e/inference.py
View File

@ -14,7 +14,7 @@ from pathlib import Path
from tqdm import tqdm, trange from tqdm import tqdm, trange
from .emb import g2p, qnt from .emb import g2p, qnt
from .emb.qnt import trim, trim_random, unload_model, repeat_extend_audio from .emb.qnt import trim, trim_random, unload_model, repeat_extend_audio, AVAILABLE_AUDIO_BACKENDS
from .emb.transcribe import transcribe from .emb.transcribe import transcribe
from .utils import to_device, set_seed, clamp, ml from .utils import to_device, set_seed, clamp, ml
@ -62,7 +62,7 @@ class TTS():
cfg.dataset.use_hdf5 = False # could use cfg.load_hdf5(), but why would it ever need to be loaded for inferencing cfg.dataset.use_hdf5 = False # could use cfg.load_hdf5(), but why would it ever need to be loaded for inferencing
# fallback to encodec if no vocos # fallback to encodec if no vocos
if cfg.audio_backend == "vocos" and not cfg.inference.audio_backends.get("vocos", False): if cfg.audio_backend == "vocos" and "vocos" not in AVAILABLE_AUDIO_BACKENDS:
_logger.warning("Vocos requested but not available, falling back to Encodec...") _logger.warning("Vocos requested but not available, falling back to Encodec...")
cfg.set_audio_backend(cfg.audio_backend) cfg.set_audio_backend(cfg.audio_backend)