yeep

app.py

@@ -37,10 +37,10 @@ def process_audio(input_text, speaker_audio, speaker_name, option_selected):
             return f"The option is not implemented yet."
         
         # Save the output audio under the speaker's name
-        speaker_output_path = f"audio/{speaker_name}.wav"
-        os.rename("audio/output.wav", speaker_output_path)
+        # speaker_output_path = f"audio/{speaker_name}.wav"
+        # os.rename("audio/output.wav", speaker_output_path)
         
-        return speaker_output_path
+        return "./tmp/audio/generated-custom.wav"
     
     except Exception as e:
         return str(e)

metaVoice.py

@@ -19,7 +19,7 @@ from fam.llm.model import GPT, GPTConfig
 from fam.llm.utils import (
     check_audio_file,
     get_default_dtype,
-    get_default_use_kv_cache,
+    # get_default_use_kv_cache,
     normalize_text,
 )
 from fam.quantiser.audio.speaker_encoder.model import SpeakerEncoder
@@ -724,9 +724,9 @@ class SamplingControllerConfig:
     init_from: str = "resume"
     """Either 'resume' (from an out_dir) or a gpt2 variant (e.g. 'gpt2-xl')."""
 
-    use_kv_cache: Optional[Literal["flash_decoding", "vanilla"]] = get_default_use_kv_cache()
-    """Type of kv caching to use for inference: 1) [none] no kv caching, 2) [flash_decoding] use the 
-    flash decoding kernel, 3) [vanilla] use torch attention with hand implemented kv-cache."""
+    # use_kv_cache: Optional[Literal["flash_decoding", "vanilla"]] = get_default_use_kv_cache()
+    # """Type of kv caching to use for inference: 1) [none] no kv caching, 2) [flash_decoding] use the 
+    # flash decoding kernel, 3) [vanilla] use torch attention with hand implemented kv-cache."""
 
     output_dir: str = "samples/"
     """Relative path to output directory"""