Update app.py
Browse files
app.py
CHANGED
@@ -1,4 +1,30 @@
|
|
1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
from nemo.collections.asr.models import ASRModel
|
3 |
import librosa
|
4 |
|
@@ -33,4 +59,4 @@ def transcribe(audio):
|
|
33 |
audio_input = gr.components.Audio()
|
34 |
|
35 |
iface = gr.Interface(transcribe, audio_input, "text", title="ASR with NeMo Canary Model")
|
36 |
-
iface.launch()
|
|
|
1 |
+
from nemo.collections.asr.models import EncDecMultiTaskModel
|
2 |
+
|
3 |
+
# Load the Canary-1B model
|
4 |
+
canary_model = EncDecMultiTaskModel.from_pretrained('nvidia/canary-1b')
|
5 |
+
|
6 |
+
# Define the input manifest file for ASR
|
7 |
+
input_manifest = {
|
8 |
+
"audio_filepath": "/path/to/audio.wav",
|
9 |
+
"duration": 1000, # duration of the audio, can be set to `None` if using NeMo main branch
|
10 |
+
"taskname": "asr",
|
11 |
+
"source_lang": "en", # language of the audio input, set `source_lang`==`target_lang` for ASR, choices=['en','de','es','fr']
|
12 |
+
"target_lang": "en", # language of the text output, choices=['en','de','es','fr']
|
13 |
+
"pnc": "yes", # whether to have PnC output, choices=['yes', 'no']
|
14 |
+
"answer": "na",
|
15 |
+
}
|
16 |
+
|
17 |
+
# Transcribe audio using the Canary-1B model
|
18 |
+
predicted_text = canary_model.transcribe(
|
19 |
+
input_manifest,
|
20 |
+
batch_size=16 # batch size to run the inference with
|
21 |
+
)
|
22 |
+
|
23 |
+
print("Predicted Text:", predicted_text)
|
24 |
+
|
25 |
+
|
26 |
+
|
27 |
+
'''import gradio as gr
|
28 |
from nemo.collections.asr.models import ASRModel
|
29 |
import librosa
|
30 |
|
|
|
59 |
audio_input = gr.components.Audio()
|
60 |
|
61 |
iface = gr.Interface(transcribe, audio_input, "text", title="ASR with NeMo Canary Model")
|
62 |
+
iface.launch()'''
|