Update app.py
Browse files
app.py
CHANGED
@@ -1,7 +1,49 @@
|
|
1 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
-
|
4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
|
6 |
-
|
7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
+
import torch
|
3 |
+
import soundfile as sf
|
4 |
+
import os
|
5 |
+
import matplotlib.pyplot as plt
|
6 |
+
import numpy as np
|
7 |
|
8 |
+
import os
|
9 |
+
import soundfile as sf
|
10 |
+
import torch
|
11 |
+
import torch.nn as nn
|
12 |
+
import torch.nn.functional as F
|
13 |
+
from torch.utils.data import Dataset, DataLoader
|
14 |
+
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor, Wav2Vec2ForSequenceClassification
|
15 |
+
from sklearn.model_selection import train_test_split
|
16 |
+
import re
|
17 |
+
from collections import Counter
|
18 |
+
from sklearn.metrics import classification_report
|
19 |
+
model = Wav2Vec2ForSequenceClassification.from_pretrained("facebook/wav2vec2-base-960h", num_labels=2).to(device)
|
20 |
+
model_path = "dysarthria_classifier12.pth"
|
21 |
+
if os.path.exists(model_path):
|
22 |
+
print(f"Loading saved model {model_path}")
|
23 |
+
model.load_state_dict(torch.load(model_path))
|
24 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
25 |
+
processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h")
|
26 |
+
def predict(file_path):
|
27 |
+
max_length = 100000
|
28 |
|
29 |
+
model.eval()
|
30 |
+
with torch.no_grad():
|
31 |
+
wav_data, _ = sf.read(file_path.name)
|
32 |
+
inputs = processor(wav_data, sampling_rate=16000, return_tensors="pt", padding=True)
|
33 |
+
|
34 |
+
input_values = inputs.input_values.squeeze(0)
|
35 |
+
if max_length - input_values.shape[-1] > 0:
|
36 |
+
input_values = torch.cat([input_values, torch.zeros((max_length - input_values.shape[-1],))], dim=-1)
|
37 |
+
else:
|
38 |
+
input_values = input_values[:max_length]
|
39 |
+
input_values = input_values.unsqueeze(0).to(device)
|
40 |
+
inputs = {"input_values": input_values}
|
41 |
+
|
42 |
+
logits = model(**inputs).logits
|
43 |
+
logits = logits.squeeze()
|
44 |
+
predicted_class_id = torch.argmax(logits, dim=-1).item()
|
45 |
+
|
46 |
+
return predicted_class_id
|
47 |
+
|
48 |
+
iface = gr.Interface(fn=predict, inputs="file", outputs="text")
|
49 |
+
iface.launch()
|