Upload folder using huggingface_hub

TTS/melo_handler.py

@@ -107,3 +107,4 @@ class MeloTTSHandler(BaseHandler):
             )
 
         self.should_listen.set()
+        yield b"END"

audio_streaming_client.py

@@ -65,12 +65,12 @@ class AudioStreamingClient:
                 time.sleep(0.1)
 
     def send_request(self, audio_data=None):
-        payload = {"inputs": ""}
+        payload = {"input_type": "speech",
+                   "inputs": ""}
 
         if audio_data is not None:
             print("Sending audio data")
             payload["inputs"] = base64.b64encode(audio_data).decode('utf-8')
-            payload["input_type"] = "speech"
 
         if self.session_id:
             payload["session_id"] = self.session_id
@@ -88,11 +88,6 @@ class AudioStreamingClient:
             if "status" in response_data and response_data["status"] == "processing":
                 print("Processing audio data")
                 self.session_state = "processing"
-            elif "status" in response_data and response_data["status"] == "completed":
-                print("Completed audio processing")
-                self.session_state = None
-                self.session_id = None
-                _ = self.send_queue.get()  # Clear the queue
 
             if "output" in response_data and response_data["output"]:
                 print("Received audio data")
@@ -104,6 +99,15 @@ class AudioStreamingClient:
                     chunk = audio_np[i:i+self.args.chunk_size]
                     self.recv_queue.put(chunk)
 
+            if "status" in response_data and response_data["status"] == "completed":
+                print("Completed audio processing")
+                self.session_state = None
+                self.session_id = None
+                while not self.recv_queue.empty():
+                    time.sleep(0.01)  # wait for the queue to empty
+                while not self.send_queue.empty():
+                    _ = self.send_queue.get()  # Clear the queue
+
         except Exception as e:
             print(f"Error sending request: {e}")
             self.session_state = "idle"  # Reset state to idle in case of error

audio_streaming_test.py

@@ -66,12 +66,12 @@ class AudioStreamingClient:
                 time.sleep(0.1)
 
     def send_request(self, audio_data=None):
-        payload = {}
+        payload = {"input_type": "speech",
+                   "inputs": ""}
 
         if audio_data is not None:
             print("Sending audio data")
             payload["inputs"] = base64.b64encode(audio_data).decode('utf-8')
-            payload["input_type"] = "speech"
 
         if self.session_id:
             payload["session_id"] = self.session_id
@@ -88,11 +88,6 @@ class AudioStreamingClient:
             if "status" in response_data and response_data["status"] == "processing":
                 print("Processing audio data")
                 self.session_state = "processing"
-            elif "status" in response_data and response_data["status"] == "completed":
-                print("Completed audio processing")
-                self.session_state = None
-                self.session_id = None
-                _ = self.send_queue.get()  # Clear the queue
 
             if "output" in response_data and response_data["output"]:
                 print("Received audio data")
@@ -104,6 +99,15 @@ class AudioStreamingClient:
                     chunk = audio_np[i:i+self.args.chunk_size]
                     self.recv_queue.put(chunk)
 
+            if "status" in response_data and response_data["status"] == "completed":
+                print("Completed audio processing")
+                self.session_state = None
+                self.session_id = None
+                while not self.recv_queue.empty():
+                    time.sleep(0.01)  # wait for the queue to empty
+                while not self.send_queue.empty():
+                    _ = self.send_queue.get()  # Clear the queue
+
         except Exception as e:
             print(f"Error sending request: {e}")
             self.session_state = "idle"  # Reset state to idle in case of error

handler.py

@@ -1,6 +1,3 @@
-import subprocess
-subprocess.run("pip install flash-attn --no-build-isolation", shell=True, check=True)
-       
 from typing import Dict, Any, List, Generator
 import torch
 import os
@@ -26,7 +23,8 @@ class EndpointHandler:
             self.parler_tts_handler_kwargs,
             self.melo_tts_handler_kwargs,
             self.chat_tts_handler_kwargs,
-        ) = get_default_arguments(mode='none', log_level='DEBUG', lm_model_name='meta-llama/Meta-Llama-3.1-8B-Instruct')
+        ) = get_default_arguments(mode='none', lm_model_name='meta-llama/Meta-Llama-3.1-8B-Instruct', tts='melo')
+
         setup_logger(self.module_kwargs.log_level)
 
         prepare_all_args(
@@ -104,6 +102,7 @@ class EndpointHandler:
             raise ValueError(f"Unsupported request type: {request_type}")
 
     def _handle_start_request(self, data: Dict[str, Any]) -> Dict[str, Any]:
+        print("Starting new session")
         session_id = str(uuid.uuid4())
         self.sessions[session_id] = {
             'status': 'new',