Spaces:
Runtime error
Runtime error
fixed bug on max output text len + adding GPU inference
Browse files
app.py
CHANGED
@@ -12,6 +12,7 @@ import torch
|
|
12 |
import gradio as gr
|
13 |
from transformers import Wav2Vec2Tokenizer, Wav2Vec2ForCTC
|
14 |
nltk.download("punkt")
|
|
|
15 |
|
16 |
|
17 |
# In[ ]:
|
@@ -22,7 +23,7 @@ model_name = "facebook/wav2vec2-base-960h"
|
|
22 |
|
23 |
#model_name = "facebook/wav2vec2-large-xlsr-53"
|
24 |
tokenizer = Wav2Vec2Tokenizer.from_pretrained(model_name)#omdel_name
|
25 |
-
model = Wav2Vec2ForCTC.from_pretrained(model_name)
|
26 |
|
27 |
|
28 |
# In[ ]:
|
@@ -81,10 +82,10 @@ def asr_transcript_long(input_file,tokenizer=tokenizer, model=model ):
|
|
81 |
# Ensure that the sample rate is 16k
|
82 |
sample_rate = librosa.get_samplerate(input_file)
|
83 |
|
84 |
-
# Stream over
|
85 |
stream = librosa.stream(
|
86 |
input_file,
|
87 |
-
block_length=
|
88 |
frame_length=sample_rate, #16000,
|
89 |
hop_length=sample_rate, #16000
|
90 |
)
|
@@ -95,15 +96,15 @@ def asr_transcript_long(input_file,tokenizer=tokenizer, model=model ):
|
|
95 |
if sample_rate !=16000:
|
96 |
speech = librosa.resample(speech, sample_rate,16000)
|
97 |
input_values = tokenizer(speech, return_tensors="pt").input_values
|
98 |
-
logits = model(input_values).logits
|
99 |
|
100 |
predicted_ids = torch.argmax(logits, dim=-1)
|
101 |
transcription = tokenizer.decode(predicted_ids[0])
|
102 |
-
#transcript +=
|
103 |
transcript += correct_casing(transcription.lower())
|
104 |
-
transcript += " "
|
105 |
|
106 |
-
return transcript
|
107 |
|
108 |
|
109 |
# In[ ]:
|
@@ -112,8 +113,8 @@ def asr_transcript_long(input_file,tokenizer=tokenizer, model=model ):
|
|
112 |
gr.Interface(asr_transcript_long,
|
113 |
#inputs = gr.inputs.Audio(source="microphone", type="filepath", optional=True, label="Please record your voice"),
|
114 |
inputs = gr.inputs.Audio(source="upload", type="filepath", optional=True, label="Upload your file here"),
|
115 |
-
outputs = gr.outputs.Textbox(label="Output Text"),
|
116 |
-
title="
|
117 |
description = "This application displays transcribed text for given audio input",
|
118 |
examples = [["Test_File1.wav"], ["Test_File2.wav"], ["Test_File3.wav"]], theme="grass").launch()
|
119 |
|
|
|
12 |
import gradio as gr
|
13 |
from transformers import Wav2Vec2Tokenizer, Wav2Vec2ForCTC
|
14 |
nltk.download("punkt")
|
15 |
+
torch_device = 'cuda'
|
16 |
|
17 |
|
18 |
# In[ ]:
|
|
|
23 |
|
24 |
#model_name = "facebook/wav2vec2-large-xlsr-53"
|
25 |
tokenizer = Wav2Vec2Tokenizer.from_pretrained(model_name)#omdel_name
|
26 |
+
model = Wav2Vec2ForCTC.from_pretrained(model_name).to(torch_device)
|
27 |
|
28 |
|
29 |
# In[ ]:
|
|
|
82 |
# Ensure that the sample rate is 16k
|
83 |
sample_rate = librosa.get_samplerate(input_file)
|
84 |
|
85 |
+
# Stream over 10 seconds chunks rather than load the full file
|
86 |
stream = librosa.stream(
|
87 |
input_file,
|
88 |
+
block_length=20, #number of seconds to split the batch
|
89 |
frame_length=sample_rate, #16000,
|
90 |
hop_length=sample_rate, #16000
|
91 |
)
|
|
|
96 |
if sample_rate !=16000:
|
97 |
speech = librosa.resample(speech, sample_rate,16000)
|
98 |
input_values = tokenizer(speech, return_tensors="pt").input_values
|
99 |
+
logits = model(input_values.to(torch_device)).logits
|
100 |
|
101 |
predicted_ids = torch.argmax(logits, dim=-1)
|
102 |
transcription = tokenizer.decode(predicted_ids[0])
|
103 |
+
#transcript += transcription.lower()
|
104 |
transcript += correct_casing(transcription.lower())
|
105 |
+
#transcript += " "
|
106 |
|
107 |
+
return transcript[:4300]
|
108 |
|
109 |
|
110 |
# In[ ]:
|
|
|
113 |
gr.Interface(asr_transcript_long,
|
114 |
#inputs = gr.inputs.Audio(source="microphone", type="filepath", optional=True, label="Please record your voice"),
|
115 |
inputs = gr.inputs.Audio(source="upload", type="filepath", optional=True, label="Upload your file here"),
|
116 |
+
outputs = gr.outputs.Textbox(type="str",label="Output Text"),
|
117 |
+
title="Transcript and Translate",
|
118 |
description = "This application displays transcribed text for given audio input",
|
119 |
examples = [["Test_File1.wav"], ["Test_File2.wav"], ["Test_File3.wav"]], theme="grass").launch()
|
120 |
|