Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -20,6 +20,7 @@ from espnet2.sds.eval.ASR_WER import handle_espnet_ASR_WER
|
|
20 |
from espnet2.sds.eval.TTS_speech_quality import TTS_psuedomos
|
21 |
from espnet2.sds.eval.LLM_Metrics import perplexity, vert, bert_score, DialoGPT_perplexity
|
22 |
from espnet2.sds.utils.chat import Chat
|
|
|
23 |
import argparse
|
24 |
|
25 |
access_token = os.environ.get("HF_TOKEN")
|
@@ -80,6 +81,7 @@ user_role = "user"
|
|
80 |
text2speech=None
|
81 |
s2t=None
|
82 |
LM_pipe=None
|
|
|
83 |
|
84 |
latency_ASR=0.0
|
85 |
latency_LM=0.0
|
@@ -144,6 +146,48 @@ def handle_eval_selection(option, TTS_audio_output, LLM_Output, ASR_audio_output
|
|
144 |
elif option=="Text Dialog Metrics":
|
145 |
yield (None,perplexity(LLM_Output.replace("\n"," "))+vert(LLM_response_arr)+bert_score(total_response_arr)+DialoGPT_perplexity(ASR_transcript.replace("\n"," "),LLM_Output.replace("\n"," ")))
|
146 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
147 |
for _ in handle_selection(TTS_name):
|
148 |
continue
|
149 |
for _ in handle_ASR_selection(ASR_name):
|
@@ -270,36 +314,44 @@ def transcribe(stream, new_chunk, option, asr_option):
|
|
270 |
stream=np.concatenate((stream,y))
|
271 |
orig_sr=sr
|
272 |
sr=16000
|
273 |
-
|
|
|
|
|
|
|
274 |
|
275 |
if array is not None:
|
276 |
print("VAD: end of speech detected")
|
277 |
start_time = time.time()
|
278 |
-
|
279 |
-
|
280 |
-
|
281 |
-
|
282 |
-
|
283 |
-
|
284 |
-
|
285 |
-
|
286 |
-
|
287 |
-
|
288 |
-
|
289 |
-
|
290 |
-
|
291 |
-
|
292 |
-
|
293 |
-
|
294 |
-
|
295 |
-
|
296 |
-
|
297 |
-
|
298 |
-
|
299 |
-
|
|
|
|
|
|
|
|
|
300 |
audio_output1=(orig_sr,stream)
|
301 |
stream=y
|
302 |
-
|
|
|
303 |
text_str1=text_str
|
304 |
if ((text_str!="") and (start_record_time is None)):
|
305 |
start_record_time=time.time()
|
@@ -338,6 +390,12 @@ with gr.Blocks(
|
|
338 |
with gr.Row():
|
339 |
with gr.Column(scale=1):
|
340 |
user_audio = gr.Audio(sources=["microphone"], streaming=True, waveform_options=gr.WaveformOptions(sample_rate=16000))
|
|
|
|
|
|
|
|
|
|
|
|
|
341 |
with gr.Row():
|
342 |
ASR_radio = gr.Radio(
|
343 |
choices=ASR_options,
|
@@ -356,6 +414,13 @@ with gr.Blocks(
|
|
356 |
label="Choose TTS:",
|
357 |
value=TTS_name,
|
358 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
359 |
with gr.Row():
|
360 |
feedback_btn = gr.Button(
|
361 |
value="Please provide your feedback after each system response below.", visible=True, interactive=False, elem_id="button"
|
@@ -391,6 +456,11 @@ with gr.Blocks(
|
|
391 |
choices=["Latency", "TTS Intelligibility", "TTS Speech Quality", "ASR WER","Text Dialog Metrics"],
|
392 |
label="Choose Evaluation metrics:",
|
393 |
)
|
|
|
|
|
|
|
|
|
|
|
394 |
output_eval_text = gr.Textbox(label="Evaluation Results")
|
395 |
state = gr.State()
|
396 |
with gr.Row():
|
@@ -421,22 +491,24 @@ with gr.Blocks(
|
|
421 |
natural_response = gr.Textbox(label="natural_response",visible=False,interactive=False)
|
422 |
diversity_response = gr.Textbox(label="diversity_response",visible=False,interactive=False)
|
423 |
ip_address = gr.Textbox(label="ip_address",visible=False,interactive=False)
|
424 |
-
callback.setup([user_audio, output_asr_text, output_text, output_audio,output_audio1,ASR_radio,LLM_radio,radio,natural_response,diversity_response,ip_address],"flagged_data_points")
|
425 |
user_audio.stream(transcribe, inputs=[state, user_audio, radio, ASR_radio], outputs=[state, output_asr_text, output_text, output_audio, output_audio1]).then(lambda *args: callback.flag(list(args)),[user_audio], None,preprocess=False)
|
426 |
radio.change(fn=handle_selection, inputs=[radio], outputs=[output_asr_text, output_text, output_audio])
|
427 |
LLM_radio.change(fn=handle_LLM_selection, inputs=[LLM_radio], outputs=[output_asr_text, output_text, output_audio])
|
428 |
ASR_radio.change(fn=handle_ASR_selection, inputs=[ASR_radio], outputs=[output_asr_text, output_text, output_audio])
|
429 |
eval_radio.change(fn=handle_eval_selection, inputs=[eval_radio,output_audio,output_text,output_audio1,output_asr_text], outputs=[eval_radio,output_eval_text])
|
|
|
|
|
430 |
output_audio.play(
|
431 |
flash_buttons, [], [natural_response,diversity_response]+btn_list
|
432 |
-
).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,ASR_radio,LLM_radio,radio], None,preprocess=False)
|
433 |
-
natural_btn1.click(natural_vote1_last_response,[],[natural_response,ip_address]+natural_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,ASR_radio,LLM_radio,radio,natural_response,diversity_response,ip_address], None,preprocess=False)
|
434 |
-
natural_btn2.click(natural_vote2_last_response,[],[natural_response,ip_address]+natural_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,ASR_radio,LLM_radio,radio,natural_response,diversity_response,ip_address], None,preprocess=False)
|
435 |
-
natural_btn3.click(natural_vote3_last_response,[],[natural_response,ip_address]+natural_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,ASR_radio,LLM_radio,radio,natural_response,diversity_response,ip_address], None,preprocess=False)
|
436 |
-
natural_btn4.click(natural_vote4_last_response,[],[natural_response,ip_address]+natural_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,ASR_radio,LLM_radio,radio,natural_response,diversity_response,ip_address], None,preprocess=False)
|
437 |
-
relevant_btn1.click(relevant_vote1_last_response,[],[diversity_response,ip_address]+relevant_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,ASR_radio,LLM_radio,radio,natural_response,diversity_response,ip_address], None,preprocess=False)
|
438 |
-
relevant_btn2.click(relevant_vote2_last_response,[],[diversity_response,ip_address]+relevant_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,ASR_radio,LLM_radio,radio,natural_response,diversity_response,ip_address], None,preprocess=False)
|
439 |
-
relevant_btn3.click(relevant_vote3_last_response,[],[diversity_response,ip_address]+relevant_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,ASR_radio,LLM_radio,radio,natural_response,diversity_response,ip_address], None,preprocess=False)
|
440 |
-
relevant_btn4.click(relevant_vote4_last_response,[],[diversity_response,ip_address]+relevant_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,ASR_radio,LLM_radio,radio,natural_response,diversity_response,ip_address], None,preprocess=False)
|
441 |
|
442 |
demo.launch(share=True)
|
|
|
20 |
from espnet2.sds.eval.TTS_speech_quality import TTS_psuedomos
|
21 |
from espnet2.sds.eval.LLM_Metrics import perplexity, vert, bert_score, DialoGPT_perplexity
|
22 |
from espnet2.sds.utils.chat import Chat
|
23 |
+
from espnet2.sds.end_to_end.mini_omni_e2e import MiniOmniE2EModel
|
24 |
import argparse
|
25 |
|
26 |
access_token = os.environ.get("HF_TOKEN")
|
|
|
81 |
text2speech=None
|
82 |
s2t=None
|
83 |
LM_pipe=None
|
84 |
+
client=None
|
85 |
|
86 |
latency_ASR=0.0
|
87 |
latency_LM=0.0
|
|
|
146 |
elif option=="Text Dialog Metrics":
|
147 |
yield (None,perplexity(LLM_Output.replace("\n"," "))+vert(LLM_response_arr)+bert_score(total_response_arr)+DialoGPT_perplexity(ASR_transcript.replace("\n"," "),LLM_Output.replace("\n"," ")))
|
148 |
|
149 |
+
def handle_eval_selection_E2E(option, TTS_audio_output, LLM_Output):
|
150 |
+
global LLM_response_arr
|
151 |
+
global total_response_arr
|
152 |
+
yield (option,gr.Textbox(visible=True))
|
153 |
+
if option=="Latency":
|
154 |
+
text=f"Total Latency: {latency_TTS:.2f}"
|
155 |
+
yield (None,text)
|
156 |
+
elif option=="TTS Intelligibility":
|
157 |
+
yield (None,handle_espnet_TTS_intelligibility(TTS_audio_output,LLM_Output))
|
158 |
+
elif option=="TTS Speech Quality":
|
159 |
+
yield (None,TTS_psuedomos(TTS_audio_output))
|
160 |
+
elif option=="Text Dialog Metrics":
|
161 |
+
yield (None,perplexity(LLM_Output.replace("\n"," "))+vert(LLM_response_arr))
|
162 |
+
|
163 |
+
def handle_type_selection(option,TTS_radio,ASR_radio,LLM_radio):
|
164 |
+
global client
|
165 |
+
global LM_pipe
|
166 |
+
global s2t
|
167 |
+
global text2speech
|
168 |
+
yield (gr.Radio(visible=False),gr.Radio(visible=False),gr.Radio(visible=False),gr.Radio(visible=False), gr.Textbox(visible=False),gr.Textbox(visible=False),gr.Audio(visible=False),gr.Radio(visible=False),gr.Radio(visible=False))
|
169 |
+
if option=="Cascaded":
|
170 |
+
client=None
|
171 |
+
for _ in handle_selection(TTS_radio):
|
172 |
+
continue
|
173 |
+
for _ in handle_ASR_selection(ASR_radio):
|
174 |
+
continue
|
175 |
+
for _ in handle_LLM_selection(LLM_radio):
|
176 |
+
continue
|
177 |
+
yield (gr.Radio(visible=True),gr.Radio(visible=True),gr.Radio(visible=True),gr.Radio(visible=False),gr.Textbox(visible=True),gr.Textbox(visible=True),gr.Audio(visible=True),gr.Radio(visible=True, interactive=True),gr.Radio(visible=False))
|
178 |
+
else:
|
179 |
+
text2speech=None
|
180 |
+
s2t=None
|
181 |
+
LM_pipe=None
|
182 |
+
handle_E2E_selection()
|
183 |
+
yield (gr.Radio(visible=False),gr.Radio(visible=False),gr.Radio(visible=False),gr.Radio(visible=True),gr.Textbox(visible=True),gr.Textbox(visible=True),gr.Audio(visible=True),gr.Radio(visible=False),gr.Radio(visible=True, interactive=True))
|
184 |
+
|
185 |
+
|
186 |
+
def handle_E2E_selection():
|
187 |
+
global client
|
188 |
+
client = MiniOmniE2EModel()
|
189 |
+
client.warmup()
|
190 |
+
|
191 |
for _ in handle_selection(TTS_name):
|
192 |
continue
|
193 |
for _ in handle_ASR_selection(ASR_name):
|
|
|
314 |
stream=np.concatenate((stream,y))
|
315 |
orig_sr=sr
|
316 |
sr=16000
|
317 |
+
if client is not None:
|
318 |
+
array=vad_model(y,orig_sr, binary=True)
|
319 |
+
else:
|
320 |
+
array=vad_model(y,orig_sr)
|
321 |
|
322 |
if array is not None:
|
323 |
print("VAD: end of speech detected")
|
324 |
start_time = time.time()
|
325 |
+
if client is not None:
|
326 |
+
(text_str, audio_output)=client(array, orig_sr)
|
327 |
+
asr_output_str=""
|
328 |
+
latency_TTS=(time.time() - start_time)
|
329 |
+
else:
|
330 |
+
prompt=s2t(array)
|
331 |
+
if len(prompt.strip().split())<2:
|
332 |
+
text_str1=text_str
|
333 |
+
yield (stream, asr_output_str, text_str1, audio_output, audio_output1)
|
334 |
+
return
|
335 |
+
|
336 |
+
|
337 |
+
asr_output_str=prompt
|
338 |
+
total_response_arr.append(prompt.replace("\n"," "))
|
339 |
+
start_LM_time=time.time()
|
340 |
+
latency_ASR=(start_LM_time - start_time)
|
341 |
+
chat.append({"role": user_role, "content": prompt})
|
342 |
+
chat_messages = chat.to_list()
|
343 |
+
generated_text = LM_pipe(chat_messages)
|
344 |
+
start_TTS_time=time.time()
|
345 |
+
latency_LM=(start_TTS_time - start_LM_time)
|
346 |
+
|
347 |
+
chat.append({"role": "assistant", "content": generated_text})
|
348 |
+
text_str=generated_text
|
349 |
+
audio_output=text2speech(text_str)
|
350 |
+
latency_TTS=(time.time() - start_TTS_time)
|
351 |
audio_output1=(orig_sr,stream)
|
352 |
stream=y
|
353 |
+
LLM_response_arr.append(text_str.replace("\n"," "))
|
354 |
+
total_response_arr.append(text_str.replace("\n"," "))
|
355 |
text_str1=text_str
|
356 |
if ((text_str!="") and (start_record_time is None)):
|
357 |
start_record_time=time.time()
|
|
|
390 |
with gr.Row():
|
391 |
with gr.Column(scale=1):
|
392 |
user_audio = gr.Audio(sources=["microphone"], streaming=True, waveform_options=gr.WaveformOptions(sample_rate=16000))
|
393 |
+
with gr.Row():
|
394 |
+
type_radio = gr.Radio(
|
395 |
+
choices=["Cascaded", "E2E"],
|
396 |
+
label="Choose type of Spoken Dialog:",
|
397 |
+
value="Cascaded",
|
398 |
+
)
|
399 |
with gr.Row():
|
400 |
ASR_radio = gr.Radio(
|
401 |
choices=ASR_options,
|
|
|
414 |
label="Choose TTS:",
|
415 |
value=TTS_name,
|
416 |
)
|
417 |
+
with gr.Row():
|
418 |
+
E2Eradio = gr.Radio(
|
419 |
+
choices=["mini-omni"],
|
420 |
+
label="Choose E2E model:",
|
421 |
+
value="mini-omni",
|
422 |
+
visible=False,
|
423 |
+
)
|
424 |
with gr.Row():
|
425 |
feedback_btn = gr.Button(
|
426 |
value="Please provide your feedback after each system response below.", visible=True, interactive=False, elem_id="button"
|
|
|
456 |
choices=["Latency", "TTS Intelligibility", "TTS Speech Quality", "ASR WER","Text Dialog Metrics"],
|
457 |
label="Choose Evaluation metrics:",
|
458 |
)
|
459 |
+
eval_radio_E2E = gr.Radio(
|
460 |
+
choices=["Latency", "TTS Intelligibility", "TTS Speech Quality","Text Dialog Metrics"],
|
461 |
+
label="Choose Evaluation metrics:",
|
462 |
+
visible=False,
|
463 |
+
)
|
464 |
output_eval_text = gr.Textbox(label="Evaluation Results")
|
465 |
state = gr.State()
|
466 |
with gr.Row():
|
|
|
491 |
natural_response = gr.Textbox(label="natural_response",visible=False,interactive=False)
|
492 |
diversity_response = gr.Textbox(label="diversity_response",visible=False,interactive=False)
|
493 |
ip_address = gr.Textbox(label="ip_address",visible=False,interactive=False)
|
494 |
+
callback.setup([user_audio, output_asr_text, output_text, output_audio,output_audio1,type_radio, ASR_radio, LLM_radio, radio, E2Eradio, natural_response,diversity_response,ip_address],"flagged_data_points")
|
495 |
user_audio.stream(transcribe, inputs=[state, user_audio, radio, ASR_radio], outputs=[state, output_asr_text, output_text, output_audio, output_audio1]).then(lambda *args: callback.flag(list(args)),[user_audio], None,preprocess=False)
|
496 |
radio.change(fn=handle_selection, inputs=[radio], outputs=[output_asr_text, output_text, output_audio])
|
497 |
LLM_radio.change(fn=handle_LLM_selection, inputs=[LLM_radio], outputs=[output_asr_text, output_text, output_audio])
|
498 |
ASR_radio.change(fn=handle_ASR_selection, inputs=[ASR_radio], outputs=[output_asr_text, output_text, output_audio])
|
499 |
eval_radio.change(fn=handle_eval_selection, inputs=[eval_radio,output_audio,output_text,output_audio1,output_asr_text], outputs=[eval_radio,output_eval_text])
|
500 |
+
eval_radio_E2E.change(fn=handle_eval_selection_E2E, inputs=[eval_radio_E2E,output_audio,output_text], outputs=[eval_radio_E2E,output_eval_text])
|
501 |
+
type_radio.change(fn=handle_type_selection,inputs=[type_radio,radio,ASR_radio,LLM_radio], outputs=[radio,ASR_radio,LLM_radio, E2Eradio,output_asr_text, output_text, output_audio,eval_radio,eval_radio_E2E])
|
502 |
output_audio.play(
|
503 |
flash_buttons, [], [natural_response,diversity_response]+btn_list
|
504 |
+
).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,type_radio, ASR_radio, LLM_radio, radio, E2Eradio], None,preprocess=False)
|
505 |
+
natural_btn1.click(natural_vote1_last_response,[],[natural_response,ip_address]+natural_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,type_radio, ASR_radio, LLM_radio, radio, E2Eradio, natural_response,diversity_response,ip_address], None,preprocess=False)
|
506 |
+
natural_btn2.click(natural_vote2_last_response,[],[natural_response,ip_address]+natural_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,type_radio, ASR_radio, LLM_radio, radio, E2Eradio, natural_response,diversity_response,ip_address], None,preprocess=False)
|
507 |
+
natural_btn3.click(natural_vote3_last_response,[],[natural_response,ip_address]+natural_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,type_radio, ASR_radio, LLM_radio, radio, E2Eradio, natural_response,diversity_response,ip_address], None,preprocess=False)
|
508 |
+
natural_btn4.click(natural_vote4_last_response,[],[natural_response,ip_address]+natural_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,type_radio, ASR_radio, LLM_radio, radio, E2Eradio, natural_response,diversity_response,ip_address], None,preprocess=False)
|
509 |
+
relevant_btn1.click(relevant_vote1_last_response,[],[diversity_response,ip_address]+relevant_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,type_radio, ASR_radio, LLM_radio, radio, E2Eradio, natural_response,diversity_response,ip_address], None,preprocess=False)
|
510 |
+
relevant_btn2.click(relevant_vote2_last_response,[],[diversity_response,ip_address]+relevant_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,type_radio, ASR_radio, LLM_radio, radio, E2Eradio, natural_response,diversity_response,ip_address], None,preprocess=False)
|
511 |
+
relevant_btn3.click(relevant_vote3_last_response,[],[diversity_response,ip_address]+relevant_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,type_radio, ASR_radio, LLM_radio, radio, E2Eradio, natural_response,diversity_response,ip_address], None,preprocess=False)
|
512 |
+
relevant_btn4.click(relevant_vote4_last_response,[],[diversity_response,ip_address]+relevant_btn_list).then(lambda *args: callback.flag(list(args)),[user_audio,output_asr_text, output_text, output_audio,output_audio1,type_radio, ASR_radio, LLM_radio, radio, E2Eradio, natural_response,diversity_response,ip_address], None,preprocess=False)
|
513 |
|
514 |
demo.launch(share=True)
|