Spaces:
Running
on
T4
Running
on
T4
add punctuator and timestamped output
Browse files- app.py +54 -18
- requirements.txt +1 -0
app.py
CHANGED
@@ -1,20 +1,28 @@
|
|
|
|
|
|
|
|
1 |
import re
|
|
|
|
|
2 |
|
3 |
import torch
|
4 |
import gradio as gr
|
5 |
import yt_dlp as youtube_dl
|
6 |
from transformers import pipeline
|
7 |
from transformers.pipelines.audio_utils import ffmpeg_read
|
|
|
8 |
|
9 |
-
import tempfile
|
10 |
-
import os
|
11 |
|
|
|
12 |
MODEL_NAME = "kotoba-tech/kotoba-whisper-v1.0"
|
13 |
BATCH_SIZE = 16
|
14 |
CHUNK_LENGTH_S = 15
|
15 |
FILE_LIMIT_MB = 1000
|
16 |
YT_LENGTH_LIMIT_S = 3600 # limit to 1 hour YouTube files
|
|
|
17 |
|
|
|
|
|
18 |
if torch.cuda.is_available():
|
19 |
torch_dtype = torch.bfloat16
|
20 |
device = "cuda:0"
|
@@ -24,6 +32,7 @@ else:
|
|
24 |
device = "cpu"
|
25 |
model_kwargs = {}
|
26 |
|
|
|
27 |
pipe = pipeline(
|
28 |
task="automatic-speech-recognition",
|
29 |
model=MODEL_NAME,
|
@@ -35,21 +44,52 @@ pipe = pipeline(
|
|
35 |
)
|
36 |
|
37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
|
39 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
if inputs is None:
|
41 |
raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")
|
42 |
-
|
43 |
-
|
44 |
-
generate_kwargs['prompt_ids'] = pipe.tokenizer.get_prompt_ids(prompt, return_tensors='pt').to(device)
|
45 |
-
text = pipe(inputs, generate_kwargs=generate_kwargs)['text']
|
46 |
-
# currently the pipeline for ASR appends the prompt at the beginning of the transcription, so remove it
|
47 |
-
return re.sub(rf"\A\s*{prompt}\s*", "", text)
|
48 |
|
49 |
def _return_yt_html_embed(yt_url):
|
50 |
video_id = yt_url.split("?v=")[-1]
|
51 |
return f'<center> <iframe width="500" height="320" src="https://www.youtube.com/embed/{video_id}"> </iframe> </center>'
|
52 |
|
|
|
53 |
def download_yt_audio(yt_url, filename):
|
54 |
info_loader = youtube_dl.YoutubeDL()
|
55 |
try:
|
@@ -76,7 +116,7 @@ def download_yt_audio(yt_url, filename):
|
|
76 |
raise gr.Error(str(err))
|
77 |
|
78 |
|
79 |
-
def yt_transcribe(yt_url, prompt,
|
80 |
html_embed_str = _return_yt_html_embed(yt_url)
|
81 |
with tempfile.TemporaryDirectory() as tmpdirname:
|
82 |
filepath = os.path.join(tmpdirname, "video.mp4")
|
@@ -85,12 +125,8 @@ def yt_transcribe(yt_url, prompt, max_filesize=75.0):
|
|
85 |
inputs = f.read()
|
86 |
inputs = ffmpeg_read(inputs, pipe.feature_extractor.sampling_rate)
|
87 |
inputs = {"array": inputs, "sampling_rate": pipe.feature_extractor.sampling_rate}
|
88 |
-
|
89 |
-
|
90 |
-
generate_kwargs['prompt_ids'] = pipe.tokenizer.get_prompt_ids(prompt, return_tensors='pt').to(device)
|
91 |
-
text = pipe(inputs, generate_kwargs=generate_kwargs)['text']
|
92 |
-
# currently the pipeline for ASR appends the prompt at the beginning of the transcription, so remove it
|
93 |
-
return html_embed_str, re.sub(rf"\A\s*{prompt}\s*", "", text)
|
94 |
|
95 |
|
96 |
demo = gr.Blocks()
|
@@ -100,7 +136,7 @@ mf_transcribe = gr.Interface(
|
|
100 |
gr.inputs.Audio(source="microphone", type="filepath", optional=True),
|
101 |
gr.inputs.Textbox(lines=1, placeholder="Prompt", optional=True)
|
102 |
],
|
103 |
-
outputs="text",
|
104 |
layout="horizontal",
|
105 |
theme="huggingface",
|
106 |
title=f"Transcribe Audio with {os.path.basename(MODEL_NAME)}",
|
@@ -114,7 +150,7 @@ file_transcribe = gr.Interface(
|
|
114 |
gr.inputs.Audio(source="upload", type="filepath", optional=True, label="Audio file"),
|
115 |
gr.inputs.Textbox(lines=1, placeholder="Prompt", optional=True)
|
116 |
],
|
117 |
-
outputs="text",
|
118 |
layout="horizontal",
|
119 |
theme="huggingface",
|
120 |
title=f"Transcribe Audio with {os.path.basename(MODEL_NAME)}",
|
|
|
1 |
+
import os
|
2 |
+
import time
|
3 |
+
import tempfile
|
4 |
import re
|
5 |
+
from math import floor
|
6 |
+
from typing import Optional
|
7 |
|
8 |
import torch
|
9 |
import gradio as gr
|
10 |
import yt_dlp as youtube_dl
|
11 |
from transformers import pipeline
|
12 |
from transformers.pipelines.audio_utils import ffmpeg_read
|
13 |
+
from punctuators.models import PunctCapSegModelONNX
|
14 |
|
|
|
|
|
15 |
|
16 |
+
# configuration
|
17 |
MODEL_NAME = "kotoba-tech/kotoba-whisper-v1.0"
|
18 |
BATCH_SIZE = 16
|
19 |
CHUNK_LENGTH_S = 15
|
20 |
FILE_LIMIT_MB = 1000
|
21 |
YT_LENGTH_LIMIT_S = 3600 # limit to 1 hour YouTube files
|
22 |
+
PUNCTUATOR = PunctCapSegModelONNX.from_pretrained("pcs_47lang")
|
23 |
|
24 |
+
|
25 |
+
# device setting
|
26 |
if torch.cuda.is_available():
|
27 |
torch_dtype = torch.bfloat16
|
28 |
device = "cuda:0"
|
|
|
32 |
device = "cpu"
|
33 |
model_kwargs = {}
|
34 |
|
35 |
+
# define the pipeline
|
36 |
pipe = pipeline(
|
37 |
task="automatic-speech-recognition",
|
38 |
model=MODEL_NAME,
|
|
|
44 |
)
|
45 |
|
46 |
|
47 |
+
def format_time(start: Optional[float], end: Optional[float]):
|
48 |
+
|
49 |
+
def _format_time(seconds: Optional[float]):
|
50 |
+
if seconds is None:
|
51 |
+
return "complete "
|
52 |
+
minutes = floor(seconds / 60)
|
53 |
+
hours = floor(seconds / 3600)
|
54 |
+
seconds = seconds - hours * 3600 - minutes * 60
|
55 |
+
m_seconds = floor(round(seconds - floor(seconds), 3) * 10 ** 3)
|
56 |
+
seconds = floor(seconds)
|
57 |
+
return f'{hours:02}:{minutes:02}:{seconds:02}.{m_seconds:03}'
|
58 |
+
|
59 |
+
return f"[{_format_time(start)}-> {_format_time(end)}]:"
|
60 |
|
61 |
+
|
62 |
+
def get_prediction(inputs, prompt: Optional[str], punctuate_text: bool = True):
|
63 |
+
generate_kwargs = {"language": "japanese", "task": "transcribe"}
|
64 |
+
if prompt:
|
65 |
+
generate_kwargs['prompt_ids'] = pipe.tokenizer.get_prompt_ids(prompt, return_tensors='pt').to(device)
|
66 |
+
prediction = pipe(inputs, return_timestamps=True, generate_kwargs=generate_kwargs)
|
67 |
+
if punctuate_text:
|
68 |
+
text_edit = PUNCTUATOR.infer([c['text'] for c in prediction['chunks']])
|
69 |
+
prediction['chunks'] = [
|
70 |
+
{
|
71 |
+
'timestamp': c['timestamp'],
|
72 |
+
'text': "".join(e) if 'unk' not in "".join(e).lower() else c['text']
|
73 |
+
} for c, e in zip(prediction['chunks'], text_edit)
|
74 |
+
]
|
75 |
+
text = "".join([c['text'] for c in prediction['chunks']])
|
76 |
+
text_timestamped = "\n".join([
|
77 |
+
f"{format_time(*c['timestamp'])} {c['text']}" for c in prediction['chunks']
|
78 |
+
])
|
79 |
+
return text, text_timestamped
|
80 |
+
|
81 |
+
|
82 |
+
def transcribe(inputs, prompt, punctuate_text: bool = True):
|
83 |
if inputs is None:
|
84 |
raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")
|
85 |
+
return get_prediction(inputs, prompt, punctuate_text)
|
86 |
+
|
|
|
|
|
|
|
|
|
87 |
|
88 |
def _return_yt_html_embed(yt_url):
|
89 |
video_id = yt_url.split("?v=")[-1]
|
90 |
return f'<center> <iframe width="500" height="320" src="https://www.youtube.com/embed/{video_id}"> </iframe> </center>'
|
91 |
|
92 |
+
|
93 |
def download_yt_audio(yt_url, filename):
|
94 |
info_loader = youtube_dl.YoutubeDL()
|
95 |
try:
|
|
|
116 |
raise gr.Error(str(err))
|
117 |
|
118 |
|
119 |
+
def yt_transcribe(yt_url, prompt, punctuate_text: bool = True):
|
120 |
html_embed_str = _return_yt_html_embed(yt_url)
|
121 |
with tempfile.TemporaryDirectory() as tmpdirname:
|
122 |
filepath = os.path.join(tmpdirname, "video.mp4")
|
|
|
125 |
inputs = f.read()
|
126 |
inputs = ffmpeg_read(inputs, pipe.feature_extractor.sampling_rate)
|
127 |
inputs = {"array": inputs, "sampling_rate": pipe.feature_extractor.sampling_rate}
|
128 |
+
text, text_timestamped = get_prediction(inputs, prompt, punctuate_text)
|
129 |
+
return html_embed_str, text, text_timestamped
|
|
|
|
|
|
|
|
|
130 |
|
131 |
|
132 |
demo = gr.Blocks()
|
|
|
136 |
gr.inputs.Audio(source="microphone", type="filepath", optional=True),
|
137 |
gr.inputs.Textbox(lines=1, placeholder="Prompt", optional=True)
|
138 |
],
|
139 |
+
outputs=["text", "text"],
|
140 |
layout="horizontal",
|
141 |
theme="huggingface",
|
142 |
title=f"Transcribe Audio with {os.path.basename(MODEL_NAME)}",
|
|
|
150 |
gr.inputs.Audio(source="upload", type="filepath", optional=True, label="Audio file"),
|
151 |
gr.inputs.Textbox(lines=1, placeholder="Prompt", optional=True)
|
152 |
],
|
153 |
+
outputs=["text", "text"],
|
154 |
layout="horizontal",
|
155 |
theme="huggingface",
|
156 |
title=f"Transcribe Audio with {os.path.basename(MODEL_NAME)}",
|
requirements.txt
CHANGED
@@ -1,3 +1,4 @@
|
|
1 |
git+https://github.com/huggingface/transformers
|
2 |
torch
|
3 |
yt-dlp
|
|
|
|
1 |
git+https://github.com/huggingface/transformers
|
2 |
torch
|
3 |
yt-dlp
|
4 |
+
punctuators
|