Update model HF download

Dockerfile

@@ -40,11 +40,11 @@ COPY --chown=user requirements.txt $HOME/app/
 
 RUN pip install -r requirements.txt
 
-RUN huggingface-cli download BSC-LT/matcha-tts-cat-onnx matcha_multispeaker_cat_opset_15_10_steps_lastwords.onnx --local-dir $HOME/app/
-
-RUN huggingface-cli download BSC-LT/vocos-mel-22khz-cat  mel_spec_22khz_cat.onnx --local-dir $HOME/app/
-
-RUN huggingface-cli download BSC-LT/vocos-mel-22khz-cat  config.yaml --local-dir $HOME/app/
+# download from inference script
+#RUN huggingface-cli download BSC-LT/matcha-tts-cat-multispeaker matcha_multispeaker_cat_opset_15_10_steps_2399.onnx --local-dir $HOME/app/
+#RUN huggingface-cli download BSC-LT/matcha-tts-cat-multispeaker config.yaml --local-dir $HOME/app/
+#RUN huggingface-cli download BSC-LT/vocos-mel-22khz-cat  mel_spec_22khz_cat.onnx --local-dir $HOME/app/
+#RUN huggingface-cli download BSC-LT/vocos-mel-22khz-cat  config.yaml --local-dir $HOME/app/
 
 COPY --chown=user . $HOME/app/
 

infer_onnx.py

@@ -10,6 +10,7 @@ import tempfile
 import yaml
 import json
 
+from huggingface_hub import hf_hub_download
 from time import perf_counter
 
 def intersperse(lst, item):
@@ -30,10 +31,10 @@ def process_text(i: int, text: str, device: torch.device):
     print(x_phones)
     return x.numpy(), x_lengths.numpy()
 
-MODEL_PATH_MATCHA_MEL="matcha_multispeaker_cat_opset_15_10_steps_lastwords.onnx"
+MODEL_PATH_MATCHA_MEL=hf_hub_download(repo_id="BSC-LT/matcha-tts-cat-multispeaker", filename="matcha_multispeaker_cat_opset_15_10_steps_2399.onnx")
 MODEL_PATH_MATCHA="matcha_hifigan_multispeaker_cat.onnx"
-MODEL_PATH_VOCOS="mel_spec_22khz_cat.onnx"
-CONFIG_PATH="config.yaml"
+MODEL_PATH_VOCOS=hf_hub_download(repo_id="BSC-LT/vocos-mel-22khz-cat", filename="mel_spec_22khz_cat.onnx")
+CONFIG_PATH=hf_hub_download(repo_id="BSC-LT/vocos-mel-22khz-cat", filename="config.yaml")
 SPEAKER_ID_DICT="spk_to_id.json"
 
 sess_options = onnxruntime.SessionOptions()