Spaces:
Runtime error
Runtime error
File size: 1,593 Bytes
25b92d1 877be96 b150df1 ed47f0e 9719ddd 7432f05 9719ddd ed47f0e b150df1 ed47f0e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 |
import pandas as pd
import gradio as gr
import torch
import torchaudio
import warnings
from cryptography.utils import CryptographyDeprecationWarning
with warnings.catch_warnings():
warnings.filterwarnings('ignore', category=CryptographyDeprecationWarning)
import paramiko
df= pd.read_csv('native_words_subset.csv')
torch._C._jit_override_can_fuse_on_cpu(False)
torch._C._jit_override_can_fuse_on_gpu(False)
torch._C._jit_set_texpr_fuser_enabled(False)
torch._C._jit_set_nvfuser_enabled(False)
loader = torch.jit.load("audio_loader.pt")
model = torch.jit.load('QuartzNet_thunderspeech_3.pt')
vocab = model.text_transform.vocab.itos
vocab[-1] = ''
def convert_probs(probs):
ids = probs.argmax(1)[0]
s = []
if vocab[ids[0]]: s.append(vocab[ids[0]])
for i in range(1,len(ids)):
if ids[i-1] != ids[i]:
new = vocab[ids[i]]
if new: s.append(new)
#return '.'.join(s)
return s
def predict(path):
audio = loader(path)
probs = model(audio, torch.tensor(audio.shape[0] * [audio.shape[-1]], device=audio.device))[0]
return convert_probs(probs)
from difflib import SequenceMatcher
def similar(a, b):
return SequenceMatcher(None, a, b).ratio()
def compare (word_choice, path):
etalon = df.loc[df['replica'] == word_choice, 'transcription'].values[0]
user = predict(path)
similar(user, etalon)
word_choice = gr.inputs.Dropdown(list(df['replica'].unique()), label="Choose a word")
gr.Interface(fn=compare, inputs=[gr.inputs.Audio(source='microphone', type='filepath', optional=True), word_choice], outputs= 'text').launch(debug=True) |