FileNotFoundError: [Errno 2] No such file or directory: 'feats_stats.npz' when using custom model path

[xunicorn]

How to reproduce:

1. docker image (Debian GNU/Linux 12 (bookworm))
2. install via your readMe
3. run this code

import os
from ukrainian_tts.tts import TTS, Voices, Stress

MODEL_PATH = "/app/models/ukrainian_tts/robinhad"
os.makedirs(MODEL_PATH, exist_ok=True)

tts = TTS(device="cpu", cache_folder=MODEL_PATH) # can try gpu, mps

UPLOAD_FOLDER = "uploads"

text = "Привіт. Раз, два, три!"

def check():
    output_path = os.path.join(UPLOAD_FOLDER, "output.wav")

    with open(output_path, mode="wb") as file:
        _, output_text = tts.tts(text, Voices.Dmytro.value, Stress.Dictionary.value, file)

    print("Accented text:", output_text)

if __name__ == "__main__":
    check()

Expectation: model will be downloaded on specific path and using it for voice generation.
Real: model was downloaded into specific folder, all files exists. But raise exception:

  state = torch.load(filename, lambda storage, loc: storage)
downloading https://github.com/robinhad/ukrainian-tts/releases/download/v6.0.0
Found /app/models/ukrainian_tts/robinhad/model.pth. Skipping download...
Found /app/models/ukrainian_tts/robinhad/config.yaml. Skipping download...
Found /app/models/ukrainian_tts/robinhad/spk_xvector.ark. Skipping download...
Found /app/models/ukrainian_tts/robinhad/feats_stats.npz. Skipping download...
downloaded.
Traceback (most recent call last):
  File "/app/src/test-tts-2.py", line 12, in <module>
    tts = TTS(device="cpu", cache_folder=MODEL_PATH) # can try gpu, mps
  File "/usr/local/lib/python3.10/site-packages/ukrainian_tts/tts.py", line 43, in __init__
  File "/usr/local/lib/python3.10/site-packages/ukrainian_tts/tts.py", line 43, in __init__
    self.__setup_cache(cache_folder)
    self.__setup_cache(cache_folder)
  File "/usr/local/lib/python3.10/site-packages/ukrainian_tts/tts.py", line 117, in __setup_cache
  File "/usr/local/lib/python3.10/site-packages/ukrainian_tts/tts.py", line 117, in __setup_cache
    self.synthesizer = Text2Speech(
  File "/usr/local/lib/python3.10/site-packages/espnet2/bin/tts_inference.py", line 92, in __init__
    model, train_args = TTSTask.build_model_from_file(
  File "/usr/local/lib/python3.10/site-packages/espnet2/bin/tts_inference.py", line 92, in __init__
    model, train_args = TTSTask.build_model_from_file(
    model, train_args = TTSTask.build_model_from_file(
  File "/usr/local/lib/python3.10/site-packages/espnet2/tasks/abs_task.py", line 1846, in build_model_from_file
    model = cls.build_model(args)
  File "/usr/local/lib/python3.10/site-packages/espnet2/tasks/tts.py", line 311, in build_model
    normalize = normalize_class(**args.normalize_conf)
  File "/usr/local/lib/python3.10/site-packages/espnet2/layers/global_mvn.py", line 40, in __init__
    stats = np.load(stats_file)
  File "/usr/local/lib/python3.10/site-packages/numpy/lib/npyio.py", line 405, in load
    fid = stack.enter_context(open(os_fspath(file), "rb"))
FileNotFoundError: [Errno 2] No such file or directory: 'feats_stats.npz'

[xunicorn]

i have tried relative path to cache folder - the same

[tdenny]

I found in the config.yaml (among the other files downloaded):

normalize_conf:
    stats_file: feats_stats.npz

By updating that line, I got it to use the cache_folder.

normalize_conf:
    stats_file: << my cache directory >>/feats_stats.npz

Maybe, some lines in the __setup_cache function to update that config line after downloading?

[Cfomodz]

Give that a shot @xunicorn and let me know any feedback you have :)