Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -9,16 +9,27 @@ import tempfile
|
|
9 |
import os
|
10 |
|
11 |
MODEL_NAME = "kotoba-tech/kotoba-whisper-v1.0"
|
12 |
-
BATCH_SIZE =
|
|
|
13 |
FILE_LIMIT_MB = 1000
|
14 |
YT_LENGTH_LIMIT_S = 3600 # limit to 1 hour YouTube files
|
15 |
|
16 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
pipe = pipeline(
|
18 |
task="automatic-speech-recognition",
|
19 |
model=MODEL_NAME,
|
20 |
-
chunk_length_s=
|
|
|
21 |
device=device,
|
|
|
22 |
)
|
23 |
|
24 |
|
@@ -26,7 +37,8 @@ pipe = pipeline(
|
|
26 |
def transcribe(inputs):
|
27 |
if inputs is None:
|
28 |
raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")
|
29 |
-
|
|
|
30 |
|
31 |
|
32 |
def _return_yt_html_embed(yt_url):
|
@@ -68,7 +80,8 @@ def yt_transcribe(yt_url, max_filesize=75.0):
|
|
68 |
inputs = f.read()
|
69 |
inputs = ffmpeg_read(inputs, pipe.feature_extractor.sampling_rate)
|
70 |
inputs = {"array": inputs, "sampling_rate": pipe.feature_extractor.sampling_rate}
|
71 |
-
|
|
|
72 |
return html_embed_str, text
|
73 |
|
74 |
|
|
|
9 |
import os
|
10 |
|
11 |
MODEL_NAME = "kotoba-tech/kotoba-whisper-v1.0"
|
12 |
+
BATCH_SIZE = 16
|
13 |
+
CHUNK_LENGTH_S = 15
|
14 |
FILE_LIMIT_MB = 1000
|
15 |
YT_LENGTH_LIMIT_S = 3600 # limit to 1 hour YouTube files
|
16 |
|
17 |
+
if torch.cuda.is_available():
|
18 |
+
torch_dtype = torch.bfloat16
|
19 |
+
device = "cuda:0"
|
20 |
+
model_kwargs = {'attn_implementation': 'sdpa'}
|
21 |
+
else:
|
22 |
+
torch_dtype = torch.float32
|
23 |
+
device = "cpu"
|
24 |
+
model_kwargs = {}
|
25 |
+
|
26 |
pipe = pipeline(
|
27 |
task="automatic-speech-recognition",
|
28 |
model=MODEL_NAME,
|
29 |
+
chunk_length_s=CHUNK_LENGTH_S,
|
30 |
+
torch_dtype=torch_dtype,
|
31 |
device=device,
|
32 |
+
model_kwargs=model_kwargs
|
33 |
)
|
34 |
|
35 |
|
|
|
37 |
def transcribe(inputs):
|
38 |
if inputs is None:
|
39 |
raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")
|
40 |
+
generate_kwargs = {"language": "japanese", "task": "transcribe"}
|
41 |
+
return pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs=generate_kwargs)["text"]
|
42 |
|
43 |
|
44 |
def _return_yt_html_embed(yt_url):
|
|
|
80 |
inputs = f.read()
|
81 |
inputs = ffmpeg_read(inputs, pipe.feature_extractor.sampling_rate)
|
82 |
inputs = {"array": inputs, "sampling_rate": pipe.feature_extractor.sampling_rate}
|
83 |
+
generate_kwargs = {"language": "japanese", "task": "transcribe"}
|
84 |
+
text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs=generate_kwargs)["text"]
|
85 |
return html_embed_str, text
|
86 |
|
87 |
|