Spaces:
Sleeping
Sleeping
Commit
·
780f607
1
Parent(s):
a355d31
Updated app.py: added HF_TOKEN and enabled gradio public url sharing
Browse files
app.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
import gradio as gr
|
2 |
from transformers import pipeline, Wav2Vec2ProcessorWithLM
|
3 |
|
@@ -5,13 +6,13 @@ def transcribe(audio, model_id):
|
|
5 |
# load processor
|
6 |
p = Wav2Vec2ProcessorWithLM.from_pretrained(model_id, revision="lm")
|
7 |
# load eval pipeline
|
8 |
-
asr = pipeline("automatic-speech-recognition", model=model_id, tokenizer=p.tokenizer, feature_extractor=p.feature_extractor, decoder=p.decoder)
|
9 |
|
10 |
text = asr(audio)["text"]
|
11 |
return text
|
12 |
|
13 |
asr_app = gr.Interface(
|
14 |
-
fn=transcribe,
|
15 |
inputs=[
|
16 |
gr.Audio(sources=["upload", "microphone"], type="filepath"),
|
17 |
gr.Dropdown(
|
@@ -21,10 +22,10 @@ asr_app = gr.Interface(
|
|
21 |
"asr-africa/wav2vec2-xls-r-1b-naijavoices-yoruba-500hr-v0"
|
22 |
]
|
23 |
)
|
24 |
-
],
|
25 |
outputs="text",
|
26 |
title="NaijaVoices ASR",
|
27 |
description="Realtime demo for Hausa, Igbo and Yoruba speech recognition using a fine-tuned Wav2Vec-XLS-R 1B model.",
|
28 |
)
|
29 |
|
30 |
-
asr_app.launch()
|
|
|
1 |
+
import os
|
2 |
import gradio as gr
|
3 |
from transformers import pipeline, Wav2Vec2ProcessorWithLM
|
4 |
|
|
|
6 |
# load processor
|
7 |
p = Wav2Vec2ProcessorWithLM.from_pretrained(model_id, revision="lm")
|
8 |
# load eval pipeline
|
9 |
+
asr = pipeline("automatic-speech-recognition", model=model_id, tokenizer=p.tokenizer, feature_extractor=p.feature_extractor, decoder=p.decoder, token=os.getenv('HF_TOKEN'))
|
10 |
|
11 |
text = asr(audio)["text"]
|
12 |
return text
|
13 |
|
14 |
asr_app = gr.Interface(
|
15 |
+
fn=transcribe,
|
16 |
inputs=[
|
17 |
gr.Audio(sources=["upload", "microphone"], type="filepath"),
|
18 |
gr.Dropdown(
|
|
|
22 |
"asr-africa/wav2vec2-xls-r-1b-naijavoices-yoruba-500hr-v0"
|
23 |
]
|
24 |
)
|
25 |
+
],
|
26 |
outputs="text",
|
27 |
title="NaijaVoices ASR",
|
28 |
description="Realtime demo for Hausa, Igbo and Yoruba speech recognition using a fine-tuned Wav2Vec-XLS-R 1B model.",
|
29 |
)
|
30 |
|
31 |
+
asr_app.launch(share=True)
|