Spaces:
Paused
Paused
makaveli10
commited on
Commit
•
795e5f6
1
Parent(s):
855f306
fix: end of segment
Browse files
whisper_live/trt_server.py
CHANGED
@@ -139,13 +139,13 @@ class TranscriptionServer:
|
|
139 |
speech_prob = self.vad_model(torch.from_numpy(frame_np.copy()), self.RATE).item()
|
140 |
if speech_prob < self.vad_threshold:
|
141 |
no_voice_activity_chunks += 1
|
142 |
-
if no_voice_activity_chunks >
|
143 |
if not self.clients[websocket].eos:
|
144 |
self.clients[websocket].set_eos(True)
|
145 |
time.sleep(0.1) # EOS stop receiving frames for a 100ms(to send output to LLM.)
|
146 |
continue
|
147 |
no_voice_activity_chunks = 0
|
148 |
-
|
149 |
|
150 |
except Exception as e:
|
151 |
logging.error(e)
|
@@ -391,13 +391,13 @@ class ServeClient:
|
|
391 |
"eos": self.eos
|
392 |
})
|
393 |
)
|
|
|
394 |
if self.eos:
|
395 |
# self.append_segment(last_segment)
|
396 |
self.timestamp_offset += duration
|
397 |
self.prompt = ' '.join(segment['text'] for segment in segments)
|
398 |
self.transcription_queue.put({"uid": self.client_uid, "prompt": self.prompt})
|
399 |
-
self.set_eos(False)
|
400 |
-
|
401 |
logging.info(f"[INFO:] Processed : {self.timestamp_offset} seconds / {self.frames_np.shape[0] / self.RATE} seconds"
|
402 |
)
|
403 |
|
@@ -435,4 +435,4 @@ class ServeClient:
|
|
435 |
"""
|
436 |
logging.info("Cleaning up.")
|
437 |
self.exit = True
|
438 |
-
self.transcriber.destroy()
|
|
|
139 |
speech_prob = self.vad_model(torch.from_numpy(frame_np.copy()), self.RATE).item()
|
140 |
if speech_prob < self.vad_threshold:
|
141 |
no_voice_activity_chunks += 1
|
142 |
+
if no_voice_activity_chunks > 3:
|
143 |
if not self.clients[websocket].eos:
|
144 |
self.clients[websocket].set_eos(True)
|
145 |
time.sleep(0.1) # EOS stop receiving frames for a 100ms(to send output to LLM.)
|
146 |
continue
|
147 |
no_voice_activity_chunks = 0
|
148 |
+
self.clients[websocket].set_eos(False)
|
149 |
|
150 |
except Exception as e:
|
151 |
logging.error(e)
|
|
|
391 |
"eos": self.eos
|
392 |
})
|
393 |
)
|
394 |
+
logging.info(f"[INFO]: {segments}, eos: {self.eos}")
|
395 |
if self.eos:
|
396 |
# self.append_segment(last_segment)
|
397 |
self.timestamp_offset += duration
|
398 |
self.prompt = ' '.join(segment['text'] for segment in segments)
|
399 |
self.transcription_queue.put({"uid": self.client_uid, "prompt": self.prompt})
|
400 |
+
# self.set_eos(False)
|
|
|
401 |
logging.info(f"[INFO:] Processed : {self.timestamp_offset} seconds / {self.frames_np.shape[0] / self.RATE} seconds"
|
402 |
)
|
403 |
|
|
|
435 |
"""
|
436 |
logging.info("Cleaning up.")
|
437 |
self.exit = True
|
438 |
+
# self.transcriber.destroy()
|