Upload folder using huggingface_hub

.gitattributes

@@ -41,6 +41,3 @@ examples/English.wav filter=lfs diff=lfs merge=lfs -text
 examples/French.wav filter=lfs diff=lfs merge=lfs -text
 examples/German.wav filter=lfs diff=lfs merge=lfs -text
 examples/Spanish.wav filter=lfs diff=lfs merge=lfs -text
-022b2161-8f56-4432-a9ae-b4bd514e4821.mp3 filter=lfs diff=lfs merge=lfs -text
-output.wav filter=lfs diff=lfs merge=lfs -text
-ref.wav filter=lfs diff=lfs merge=lfs -text

app.py

@@ -313,4 +313,4 @@ if __name__ == "__main__":
     inference_fct = get_inference_wrapper(inference_engine)
 
     app = build_app(inference_fct, args.theme)
-    app.queue(api_open=True).launch(show_error=True, show_api=True)
+    app.queue(api_open=True).launch(show_error=True, show_api=True, server_name="0.0.0.0", server_port=18888)

fish_speech/models/text2semantic/llama.py

@@ -48,7 +48,7 @@ class BaseModelArgs:
 
     # Codebook configs
     codebook_size: int = 160
-    num_codebooks: int = 9
+    num_codebooks: int = 4
 
     # Gradient checkpointing
     use_gradient_checkpointing: bool = True

tools/api.py

@@ -136,7 +136,7 @@ async def other_exception_handler(exc: "Exception"):
 
 
 def load_audio(reference_audio, sr):
-    if len(str(reference_audio)) > 255 or not Path(reference_audio).exists():
+    if len(reference_audio) > 255 or not Path(reference_audio).exists():
         audio_data = reference_audio
         reference_audio = io.BytesIO(audio_data)
 

tools/vqgan/inference.py

@@ -14,7 +14,7 @@ from omegaconf import OmegaConf
 from tools.file import AUDIO_EXTENSIONS
 
 # register eval resolver
-#OmegaConf.register_new_resolver("eval", eval)
+OmegaConf.register_new_resolver("eval", eval)
 
 
 def load_model(config_name, checkpoint_path, device="cuda"):