Fix issues for cloning voice feature (#562)

Signed-off-by: Lv, Liang1 <liang1.lv@intel.com>

intel_extension_for_transformers/neural_chat/examples/deployment/talkingbot/server/backend/talkingbot.yaml

@@ -20,7 +20,7 @@
 #################################################################################
 #                             SERVER SETTING                                    #
 #################################################################################
-host: 127.0.0.1
+host: 0.0.0.0
 port: 8888
 
 model_name_or_path: "Intel/neural-chat-7b-v1-1"

intel_extension_for_transformers/neural_chat/examples/deployment/textbot/backend/textbot.yaml

@@ -20,7 +20,7 @@
 #################################################################################
 #                             SERVER SETTING                                    #
 #################################################################################
-host: 127.0.0.1
+host: 0.0.0.0
 port: 8000
 
 model_name_or_path: "starcoder_int8"

intel_extension_for_transformers/neural_chat/pipeline/plugins/audio/tts.py

@@ -201,6 +201,8 @@ def stream_text2speech(self, generator, output_audio_path, voice="default"):
 
 
     def post_llm_inference_actions(self, text_or_generator):
+        from intel_extension_for_transformers.neural_chat.plugins import plugins
+        self.voice = plugins.tts.args["voice"]
         if self.stream_mode:
             def cache_words_into_sentences():
                 buffered_texts = []

intel_extension_for_transformers/neural_chat/server/restful/voicechat_api.py

@@ -22,6 +22,7 @@
 from fastapi import File, UploadFile, Form
 from pydub import AudioSegment
 from ...config import GenerationConfig
+from ...plugins import plugins
 import base64
 import torch
 
@@ -47,9 +48,10 @@ def handle_voice_asr_request(self, filename: str) -> str:
         except Exception as e:
             raise Exception(e)
 
-    async def handle_voice_chat_request(self, prompt: str, audio_output_path: Optional[str]=None) -> str:
+    async def handle_voice_chat_request(self, prompt: str, voice: str, audio_output_path: Optional[str]=None) -> str:
         chatbot = self.get_chatbot()
         try:
+            plugins.tts.args["voice"] = voice
             config = GenerationConfig(audio_output_path=audio_output_path)
             result, link = chatbot.chat_stream(query=prompt, config=config)
             def audio_file_generate(result):
@@ -90,6 +92,7 @@ async def handle_talkingbot_asr(file: UploadFile = File(...)):
     audio = AudioSegment.from_file("tmp_audio_bytes")
     audio = audio.set_frame_rate(16000)
     # bytes to wav
+    file_name = file_name +'.wav'
     audio.export(f"{file_name}", format="wav")
     asr_result = router.handle_voice_asr_request(file_name)
     return {"asr_result": asr_result}
@@ -105,7 +108,7 @@ async def talkingbot(request: Request):
 
     logger.info(f'Received prompt: {text}, and use voice: {voice} knowledge_id: {knowledge_id}')
 
-    return await router.handle_voice_chat_request(text, audio_output_path)
+    return await router.handle_voice_chat_request(text, voice, audio_output_path)
 
 @router.post("/v1/talkingbot/create_embedding")
 async def create_speaker_embedding(file: UploadFile = File(...)):
@@ -120,5 +123,5 @@ async def create_speaker_embedding(file: UploadFile = File(...)):
     audio = AudioSegment.from_file(f"tmp_spk_{file_name}")
     audio.export(f"{spk_id}", format="mp3")
 
-    router.handle_create_speaker_embedding(spk_id)
+    await router.handle_create_speaker_embedding(spk_id)
     return {"spk_id": spk_id}