Spaces:
Running
on
T4
Running
on
T4
Anna Sun
commited on
Commit
•
c1e0588
1
Parent(s):
7fb1760
Model fixes
Browse files- app.py +102 -146
- models/vad_s2st_sc_24khz_main.yaml +24 -0
- simuleval_transcoder.py +18 -49
app.py
CHANGED
@@ -1,24 +1,15 @@
|
|
1 |
from __future__ import annotations
|
2 |
|
3 |
-
import os
|
4 |
-
|
5 |
import gradio as gr
|
6 |
import numpy as np
|
7 |
-
import torch
|
8 |
-
import torchaudio
|
9 |
-
import sys
|
10 |
-
from sample_wav import sample_wav
|
11 |
-
np.set_printoptions(threshold=sys.maxsize)
|
12 |
|
13 |
-
|
|
|
14 |
|
15 |
-
from pydub import AudioSegment
|
16 |
import time
|
17 |
-
from
|
|
|
18 |
|
19 |
-
from seamless_communication.cli.streaming.agents.tt_waitk_unity_s2t_m4t import (
|
20 |
-
TestTimeWaitKUnityS2TM4T,
|
21 |
-
)
|
22 |
|
23 |
language_code_to_name = {
|
24 |
"cmn": "Mandarin Chinese",
|
@@ -32,7 +23,17 @@ LANGUAGE_NAME_TO_CODE = {v: k for k, v in language_code_to_name.items()}
|
|
32 |
|
33 |
DEFAULT_TARGET_LANGUAGE = "English"
|
34 |
|
35 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
36 |
transcoder = SimulevalTranscoder(
|
37 |
sample_rate=48_000,
|
38 |
debug=False,
|
@@ -41,93 +42,97 @@ transcoder = SimulevalTranscoder(
|
|
41 |
|
42 |
def start_recording():
|
43 |
logger.debug(f"start_recording: starting transcoder")
|
|
|
44 |
transcoder.start()
|
|
|
45 |
|
|
|
|
|
46 |
|
47 |
-
|
48 |
-
|
49 |
-
|
|
|
50 |
|
51 |
-
# print(sample_rate)
|
52 |
-
# print("--------- start \n")
|
53 |
-
# # print(data)
|
54 |
-
# def map(x):
|
55 |
-
# return x
|
56 |
-
# print(data.tolist())
|
57 |
-
# print("--------- end \n")
|
58 |
|
|
|
59 |
|
|
|
|
|
|
|
60 |
transcoder.process_incoming_bytes(data.tobytes(), 'eng', sample_rate)
|
|
|
|
|
|
|
|
|
|
|
61 |
|
62 |
speech_and_text_output = transcoder.get_buffered_output()
|
63 |
if speech_and_text_output is None:
|
64 |
logger.debug("No output from transcoder.get_buffered_output()")
|
65 |
-
return None, None
|
66 |
|
67 |
-
logger.debug(f"We DID get output from the transcoder!
|
68 |
|
69 |
text = None
|
70 |
speech = None
|
71 |
|
72 |
if speech_and_text_output.speech_samples:
|
73 |
-
speech = (speech_and_text_output.
|
74 |
|
75 |
if speech_and_text_output.text:
|
76 |
text = speech_and_text_output.text
|
77 |
if speech_and_text_output.final:
|
78 |
text += "\n"
|
79 |
|
80 |
-
return speech, text
|
81 |
-
|
82 |
-
def
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
audio_np_array = audio_bytes
|
102 |
-
|
103 |
-
|
104 |
-
# combine translated wav
|
105 |
-
if type(translated_audio_bytes_state) is not tuple:
|
106 |
-
translated_audio_bytes_state = (sample_rate, audio_np_array)
|
107 |
-
# translated_audio_bytes_state = np.array([])
|
108 |
else:
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
|
|
|
|
|
|
|
|
|
131 |
|
132 |
def clear():
|
133 |
logger.debug(f"Clearing State")
|
@@ -138,105 +143,56 @@ def blocks():
|
|
138 |
with gr.Blocks() as demo:
|
139 |
|
140 |
with gr.Row():
|
141 |
-
#
|
142 |
target_language = gr.Dropdown(
|
143 |
label="Target language",
|
144 |
choices=S2ST_TARGET_LANGUAGE_NAMES,
|
145 |
value=DEFAULT_TARGET_LANGUAGE,
|
146 |
)
|
147 |
|
148 |
-
translated_audio_bytes_state = gr.State(None)
|
149 |
translated_text_state = gr.State("")
|
150 |
|
151 |
input_audio = gr.Audio(
|
152 |
label="Input Audio",
|
153 |
-
|
154 |
-
sources=["microphone"], # new gradio seems to call this less often...
|
155 |
streaming=True,
|
156 |
)
|
157 |
|
158 |
-
# input_audio = gr.Audio(
|
159 |
-
# label="Input Audio",
|
160 |
-
# type="filepath",
|
161 |
-
# source="microphone",
|
162 |
-
# streaming=True,
|
163 |
-
# )
|
164 |
-
|
165 |
-
most_recent_input_audio_segment = gr.Audio(
|
166 |
-
label="Recent Input Audio Segment segments",
|
167 |
-
# format="bytes",
|
168 |
-
streaming=True
|
169 |
-
)
|
170 |
-
|
171 |
-
# Force translate
|
172 |
-
stream_as_bytes_btn = gr.Button("Force translate most recent recording segment (ask for model output)")
|
173 |
output_translation_segment = gr.Audio(
|
174 |
label="Translated audio segment",
|
175 |
-
autoplay=
|
176 |
-
streaming=True,
|
177 |
-
type="numpy",
|
178 |
-
)
|
179 |
-
|
180 |
-
output_translation_combined = gr.Audio(
|
181 |
-
label="Translated audio combined",
|
182 |
-
autoplay=False,
|
183 |
streaming=True,
|
184 |
-
type="numpy",
|
185 |
)
|
186 |
|
187 |
-
#
|
188 |
stream_output_text = gr.Textbox(label="Translated text")
|
189 |
|
190 |
-
|
191 |
-
|
192 |
-
[input_audio, translated_audio_bytes_state, translated_text_state],
|
193 |
-
[
|
194 |
-
most_recent_input_audio_segment,
|
195 |
-
output_translation_segment,
|
196 |
-
output_translation_combined,
|
197 |
-
stream_output_text,
|
198 |
-
translated_audio_bytes_state,
|
199 |
-
translated_text_state,
|
200 |
-
],
|
201 |
)
|
202 |
-
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
|
208 |
-
|
209 |
-
|
210 |
-
# stream_output_text,
|
211 |
-
# translated_audio_bytes_state,
|
212 |
-
# translated_text_state,
|
213 |
-
# ],
|
214 |
-
# )
|
215 |
-
|
216 |
-
input_audio.stream(
|
217 |
streaming_input_callback,
|
218 |
-
|
219 |
[
|
220 |
-
most_recent_input_audio_segment,
|
221 |
output_translation_segment,
|
222 |
-
output_translation_combined,
|
223 |
stream_output_text,
|
224 |
-
translated_audio_bytes_state,
|
225 |
translated_text_state,
|
226 |
],
|
227 |
)
|
228 |
-
|
229 |
-
|
230 |
-
start_recording,
|
231 |
-
)
|
232 |
-
|
233 |
-
input_audio.clear(
|
234 |
-
clear, None, [translated_audio_bytes_state, translated_text_state]
|
235 |
)
|
236 |
-
input_audio.
|
237 |
-
|
238 |
)
|
239 |
|
240 |
-
demo.
|
241 |
|
242 |
blocks()
|
|
|
1 |
from __future__ import annotations
|
2 |
|
|
|
|
|
3 |
import gradio as gr
|
4 |
import numpy as np
|
|
|
|
|
|
|
|
|
|
|
5 |
|
6 |
+
import asyncio
|
7 |
+
from simuleval_transcoder import SimulevalTranscoder, logger
|
8 |
|
|
|
9 |
import time
|
10 |
+
from simuleval.utils.agent import build_system_from_dir
|
11 |
+
import torch
|
12 |
|
|
|
|
|
|
|
13 |
|
14 |
language_code_to_name = {
|
15 |
"cmn": "Mandarin Chinese",
|
|
|
23 |
|
24 |
DEFAULT_TARGET_LANGUAGE = "English"
|
25 |
|
26 |
+
|
27 |
+
def build_agent(model_path, config_name=None):
|
28 |
+
agent = build_system_from_dir(
|
29 |
+
model_path, config_name=config_name,
|
30 |
+
)
|
31 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
32 |
+
agent.to(device, fp16=True)
|
33 |
+
|
34 |
+
return agent
|
35 |
+
|
36 |
+
agent = build_agent("models", "vad_s2st_sc_24khz_main.yaml")
|
37 |
transcoder = SimulevalTranscoder(
|
38 |
sample_rate=48_000,
|
39 |
debug=False,
|
|
|
42 |
|
43 |
def start_recording():
|
44 |
logger.debug(f"start_recording: starting transcoder")
|
45 |
+
transcoder.reset_states()
|
46 |
transcoder.start()
|
47 |
+
transcoder.close = False
|
48 |
|
49 |
+
def stop_recording():
|
50 |
+
transcoder.close = True
|
51 |
|
52 |
+
class MyState:
|
53 |
+
def __init__(self):
|
54 |
+
self.queue = asyncio.Queue()
|
55 |
+
self.close = False
|
56 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
57 |
|
58 |
+
s = MyState()
|
59 |
|
60 |
+
def process_incoming_bytes(audio):
|
61 |
+
logger.debug(f"process_bytes: incoming audio")
|
62 |
+
sample_rate, data = audio
|
63 |
transcoder.process_incoming_bytes(data.tobytes(), 'eng', sample_rate)
|
64 |
+
s.queue.put_nowait(audio)
|
65 |
+
|
66 |
+
|
67 |
+
|
68 |
+
def get_buffered_output():
|
69 |
|
70 |
speech_and_text_output = transcoder.get_buffered_output()
|
71 |
if speech_and_text_output is None:
|
72 |
logger.debug("No output from transcoder.get_buffered_output()")
|
73 |
+
return None, None, None
|
74 |
|
75 |
+
logger.debug(f"We DID get output from the transcoder!")
|
76 |
|
77 |
text = None
|
78 |
speech = None
|
79 |
|
80 |
if speech_and_text_output.speech_samples:
|
81 |
+
speech = (speech_and_text_output.speech_sample_rate, speech_and_text_output.speech_samples)
|
82 |
|
83 |
if speech_and_text_output.text:
|
84 |
text = speech_and_text_output.text
|
85 |
if speech_and_text_output.final:
|
86 |
text += "\n"
|
87 |
|
88 |
+
return speech, text, speech_and_text_output.final
|
89 |
+
|
90 |
+
def streaming_input_callback():
|
91 |
+
final = False
|
92 |
+
max_wait_s = 15
|
93 |
+
wait_s = 0
|
94 |
+
translated_text_state = ""
|
95 |
+
while not transcoder.close:
|
96 |
+
translated_wav_segment, translated_text, final = get_buffered_output()
|
97 |
+
|
98 |
+
if translated_wav_segment is None and translated_text is None:
|
99 |
+
time.sleep(0.3)
|
100 |
+
wait_s += 0.3
|
101 |
+
if wait_s >= max_wait_s:
|
102 |
+
transcoder.close = True
|
103 |
+
continue
|
104 |
+
wait_s = 0
|
105 |
+
if translated_wav_segment is not None:
|
106 |
+
sample_rate, audio_bytes = translated_wav_segment
|
107 |
+
print("output sample rate", sample_rate)
|
108 |
+
translated_wav_segment = sample_rate, np.array(audio_bytes)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
109 |
else:
|
110 |
+
translated_wav_segment = bytes()
|
111 |
+
|
112 |
+
if translated_text is not None:
|
113 |
+
translated_text_state += " | " + str(translated_text)
|
114 |
+
|
115 |
+
stream_output_text = translated_text_state
|
116 |
+
if translated_text is not None:
|
117 |
+
print("translated:", translated_text_state)
|
118 |
+
yield [
|
119 |
+
translated_wav_segment,
|
120 |
+
stream_output_text,
|
121 |
+
translated_text_state,
|
122 |
+
]
|
123 |
+
|
124 |
+
|
125 |
+
def streaming_callback_dummy():
|
126 |
+
while not transcoder.close:
|
127 |
+
if s.queue.empty():
|
128 |
+
print("empty")
|
129 |
+
yield bytes()
|
130 |
+
time.sleep(0.3)
|
131 |
+
else:
|
132 |
+
print("audio")
|
133 |
+
audio = s.queue.get_nowait()
|
134 |
+
s.queue.task_done()
|
135 |
+
yield audio
|
136 |
|
137 |
def clear():
|
138 |
logger.debug(f"Clearing State")
|
|
|
143 |
with gr.Blocks() as demo:
|
144 |
|
145 |
with gr.Row():
|
146 |
+
# TODO: add target language switching
|
147 |
target_language = gr.Dropdown(
|
148 |
label="Target language",
|
149 |
choices=S2ST_TARGET_LANGUAGE_NAMES,
|
150 |
value=DEFAULT_TARGET_LANGUAGE,
|
151 |
)
|
152 |
|
|
|
153 |
translated_text_state = gr.State("")
|
154 |
|
155 |
input_audio = gr.Audio(
|
156 |
label="Input Audio",
|
157 |
+
sources=["microphone"],
|
|
|
158 |
streaming=True,
|
159 |
)
|
160 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
161 |
output_translation_segment = gr.Audio(
|
162 |
label="Translated audio segment",
|
163 |
+
autoplay=True,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
164 |
streaming=True,
|
|
|
165 |
)
|
166 |
|
167 |
+
# Output text segment
|
168 |
stream_output_text = gr.Textbox(label="Translated text")
|
169 |
|
170 |
+
input_audio.clear(
|
171 |
+
clear, None, [output_translation_segment, translated_text_state]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
172 |
)
|
173 |
+
input_audio.start_recording(
|
174 |
+
clear, None, [output_translation_segment, translated_text_state]
|
175 |
+
).then(
|
176 |
+
start_recording
|
177 |
+
).then(
|
178 |
+
# streaming_callback_dummy, # TODO: autoplay works fine with streaming_callback_dummy
|
179 |
+
# None,
|
180 |
+
# output_translation_segment
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
181 |
streaming_input_callback,
|
182 |
+
None,
|
183 |
[
|
|
|
184 |
output_translation_segment,
|
|
|
185 |
stream_output_text,
|
|
|
186 |
translated_text_state,
|
187 |
],
|
188 |
)
|
189 |
+
input_audio.stop_recording(
|
190 |
+
stop_recording
|
|
|
|
|
|
|
|
|
|
|
191 |
)
|
192 |
+
input_audio.stream(
|
193 |
+
process_incoming_bytes, [input_audio], None
|
194 |
)
|
195 |
|
196 |
+
demo.launch(server_port=6010)
|
197 |
|
198 |
blocks()
|
models/vad_s2st_sc_24khz_main.yaml
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
agent_class: seamless_communication.streaming.agents.mma_m4t_s2st.SeamlessS2STJointVADAgent
|
2 |
+
# checkpoint: checkpoint_best.pt
|
3 |
+
monotonic_decoder_model_name: seamless_streaming_monotonic_decoder
|
4 |
+
unity_model_name: seamless_streaming_unity
|
5 |
+
sentencepiece_model: spm_256k_nllb100.model
|
6 |
+
|
7 |
+
task: s2st
|
8 |
+
tgt_lang: "eng"
|
9 |
+
min_unit_chunk_size: 50
|
10 |
+
decision_threshold: 0.7
|
11 |
+
no_early_stop: True
|
12 |
+
block_ngrams: True
|
13 |
+
vocoder_name: vocoder_pretssel
|
14 |
+
wav2vec_yaml: wav2vec.yaml
|
15 |
+
# min_starting_wait: 12
|
16 |
+
# min_starting_wait_w2vbert: 192
|
17 |
+
|
18 |
+
config_yaml: cfg_fbank_u2t.yaml
|
19 |
+
vocoder_sample_rate: 24000
|
20 |
+
upstream_idx: 1
|
21 |
+
detokenize_only: True
|
22 |
+
device: cuda:0
|
23 |
+
max_len_a: 0
|
24 |
+
max_len_b: 1000
|
simuleval_transcoder.py
CHANGED
@@ -20,13 +20,6 @@ import time
|
|
20 |
import random
|
21 |
import colorlog
|
22 |
|
23 |
-
# Sanity check that pipeline is loadable
|
24 |
-
from seamless_communication.cli.streaming.agents.tt_waitk_unity_s2t_m4t import (
|
25 |
-
# TestTimeWaitKUnityS2TM4T,
|
26 |
-
TestTimeWaitKUnityS2TM4TVAD
|
27 |
-
)
|
28 |
-
|
29 |
-
from simuleval.utils.agent import build_system_args
|
30 |
|
31 |
MODEL_SAMPLE_RATE = 16_000
|
32 |
|
@@ -49,35 +42,6 @@ logger.addHandler(handler)
|
|
49 |
logger.setLevel(logging.DEBUG)
|
50 |
|
51 |
|
52 |
-
# TODO: Integrate this better so target lang and others can be changed. Also currently dependent on devserver internals
|
53 |
-
def build_agent():
|
54 |
-
config = {
|
55 |
-
'dataloader': 'fairseq2_s2t',
|
56 |
-
'data_file': '/large_experiments/seamless/ust/abinesh/data/s2st50_manifests/50-10/simuleval/dev_mtedx_filt_50-10_debug.tsv',
|
57 |
-
'model_name': 'seamlessM4T_v2_large',
|
58 |
-
'device': 'cuda:0',
|
59 |
-
'source_segment_size': 320,
|
60 |
-
'waitk_lagging': 7,
|
61 |
-
'fixed_pre_decision_ratio': 2,
|
62 |
-
'init_target_tokens': '</s> __eng__',
|
63 |
-
'max_len_a': 0,
|
64 |
-
'max_len_b': 200,
|
65 |
-
'agent_class': 'seamless_communication.cli.streaming.agents.tt_waitk_unity_s2t_m4t.TestTimeWaitKUnityS2TM4TVAD',
|
66 |
-
'task': 's2st',
|
67 |
-
'tgt_lang': 'eng',
|
68 |
-
'latency_metrics': 'StartOffset EndOffset AL',
|
69 |
-
'output': 'TestTimeWaitKUnityS2TM4TVAD-wait7-debug'
|
70 |
-
}
|
71 |
-
|
72 |
-
agent , _ = build_system_args(config)
|
73 |
-
# device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
74 |
-
# agent.to(device, fp16=True)
|
75 |
-
logger.info(
|
76 |
-
f"Successfully built simuleval agent"
|
77 |
-
)
|
78 |
-
|
79 |
-
return agent
|
80 |
-
|
81 |
class SpeechAndTextOutput:
|
82 |
def __init__(
|
83 |
self,
|
@@ -150,7 +114,7 @@ class OutputSegments:
|
|
150 |
for segment in segment_list:
|
151 |
speech_out += segment.content
|
152 |
output.speech_samples = speech_out
|
153 |
-
output.speech_sample_rate =
|
154 |
elif isinstance(segment_list[0], EmptySegment):
|
155 |
continue
|
156 |
else:
|
@@ -212,8 +176,9 @@ def convert_waveform(
|
|
212 |
return waveform, sample_rate
|
213 |
|
214 |
class SimulevalTranscoder:
|
215 |
-
def __init__(self, sample_rate, debug, buffer_limit):
|
216 |
-
|
|
|
217 |
self.input_queue = asyncio.Queue()
|
218 |
self.output_queue = asyncio.Queue()
|
219 |
self.states = self.agent.build_states()
|
@@ -289,6 +254,7 @@ class SimulevalTranscoder:
|
|
289 |
)
|
290 |
# # segment is array([0, 0, 0, ..., 0, 0, 0], dtype=int16)
|
291 |
self.input_queue.put_nowait(segment)
|
|
|
292 |
|
293 |
def get_input_segment(self):
|
294 |
if self.input_queue.empty():
|
@@ -340,10 +306,11 @@ class SimulevalTranscoder:
|
|
340 |
self.first_input_ts = self.get_states_root().first_input_ts
|
341 |
|
342 |
if not output_segment.is_empty:
|
|
|
343 |
self.output_queue.put_nowait(output_segment)
|
344 |
|
345 |
if output_segment.finished:
|
346 |
-
|
347 |
|
348 |
self.reset_states()
|
349 |
|
@@ -360,17 +327,19 @@ class SimulevalTranscoder:
|
|
360 |
if self.close:
|
361 |
return # closes the thread
|
362 |
|
363 |
-
|
364 |
while not self.close:
|
365 |
input_segment = self.get_input_segment()
|
366 |
if input_segment is None:
|
367 |
-
|
368 |
-
|
369 |
-
|
370 |
-
|
|
|
371 |
continue
|
|
|
372 |
self.process_pipeline_impl(input_segment)
|
373 |
-
|
374 |
|
375 |
def process_pipeline_once(self):
|
376 |
if self.close:
|
@@ -392,7 +361,7 @@ class SimulevalTranscoder:
|
|
392 |
return output_chunk
|
393 |
|
394 |
def start(self):
|
395 |
-
|
396 |
threading.Thread(target=self.process_pipeline_loop).start()
|
397 |
|
398 |
def first_translation_time(self):
|
@@ -400,7 +369,7 @@ class SimulevalTranscoder:
|
|
400 |
|
401 |
def get_buffered_output(self) -> SpeechAndTextOutput:
|
402 |
now = time.time() * 1000
|
403 |
-
|
404 |
while not self.output_queue.empty():
|
405 |
tmp_out = self.get_output_segment()
|
406 |
if tmp_out and tmp_out.compute_length(self.g2p) > 0:
|
@@ -452,4 +421,4 @@ class SimulevalTranscoder:
|
|
452 |
self.output_buffer.append(segment.segments)
|
453 |
|
454 |
def _compute_phoneme_count(self, string: str) -> int:
|
455 |
-
return len([x for x in self.g2p(string) if x != " "])
|
|
|
20 |
import random
|
21 |
import colorlog
|
22 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
|
24 |
MODEL_SAMPLE_RATE = 16_000
|
25 |
|
|
|
42 |
logger.setLevel(logging.DEBUG)
|
43 |
|
44 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
45 |
class SpeechAndTextOutput:
|
46 |
def __init__(
|
47 |
self,
|
|
|
114 |
for segment in segment_list:
|
115 |
speech_out += segment.content
|
116 |
output.speech_samples = speech_out
|
117 |
+
output.speech_sample_rate = segment.sample_rate
|
118 |
elif isinstance(segment_list[0], EmptySegment):
|
119 |
continue
|
120 |
else:
|
|
|
176 |
return waveform, sample_rate
|
177 |
|
178 |
class SimulevalTranscoder:
|
179 |
+
def __init__(self, agent, sample_rate, debug, buffer_limit):
|
180 |
+
# agent is stateless
|
181 |
+
self.agent = agent
|
182 |
self.input_queue = asyncio.Queue()
|
183 |
self.output_queue = asyncio.Queue()
|
184 |
self.states = self.agent.build_states()
|
|
|
254 |
)
|
255 |
# # segment is array([0, 0, 0, ..., 0, 0, 0], dtype=int16)
|
256 |
self.input_queue.put_nowait(segment)
|
257 |
+
print("process_incoming: put input_queue")
|
258 |
|
259 |
def get_input_segment(self):
|
260 |
if self.input_queue.empty():
|
|
|
306 |
self.first_input_ts = self.get_states_root().first_input_ts
|
307 |
|
308 |
if not output_segment.is_empty:
|
309 |
+
print("PUT IN OUTPUT QUEUE")
|
310 |
self.output_queue.put_nowait(output_segment)
|
311 |
|
312 |
if output_segment.finished:
|
313 |
+
print("OUTPUT SEGMENT IS FINISHED. Resetting states.")
|
314 |
|
315 |
self.reset_states()
|
316 |
|
|
|
327 |
if self.close:
|
328 |
return # closes the thread
|
329 |
|
330 |
+
print("processing_pipeline")
|
331 |
while not self.close:
|
332 |
input_segment = self.get_input_segment()
|
333 |
if input_segment is None:
|
334 |
+
if self.get_states_root().is_fresh_state: # TODO: this is hacky
|
335 |
+
time.sleep(0.3)
|
336 |
+
print("loop: input_queue empty")
|
337 |
+
else:
|
338 |
+
time.sleep(0.03)
|
339 |
continue
|
340 |
+
print("loop: got input_segment")
|
341 |
self.process_pipeline_impl(input_segment)
|
342 |
+
print("finished processing_pipeline")
|
343 |
|
344 |
def process_pipeline_once(self):
|
345 |
if self.close:
|
|
|
361 |
return output_chunk
|
362 |
|
363 |
def start(self):
|
364 |
+
print("starting transcoder in a thread")
|
365 |
threading.Thread(target=self.process_pipeline_loop).start()
|
366 |
|
367 |
def first_translation_time(self):
|
|
|
369 |
|
370 |
def get_buffered_output(self) -> SpeechAndTextOutput:
|
371 |
now = time.time() * 1000
|
372 |
+
print(f"get_buffered_output queue size: {self.output_queue.qsize()}")
|
373 |
while not self.output_queue.empty():
|
374 |
tmp_out = self.get_output_segment()
|
375 |
if tmp_out and tmp_out.compute_length(self.g2p) > 0:
|
|
|
421 |
self.output_buffer.append(segment.segments)
|
422 |
|
423 |
def _compute_phoneme_count(self, string: str) -> int:
|
424 |
+
return len([x for x in self.g2p(string) if x != " "])
|