AI Research Engineer in ReturnZero
Neural speech recognition, Neural speech synthesis, MLOps, Deep-learning, Machine-learning
PyTorch implementation of "ContextNet: Improving Convolutional Neural Networks for Automatic Speech Recognition with Global Context" (INTERSPEECH 2020)
License: Apache License 2.0
AI Research Engineer in ReturnZero
Neural speech recognition, Neural speech synthesis, MLOps, Deep-learning, Machine-learning
I would like to ask what input size can be understood as a voice signal
ContextNet decode function in model.py
@torch.no_grad()
def decode(self, encoder_output: Tensor, max_lengths: int) -> Tensor:
r"""
Decode `encoder_output`.
Args:
**encoder_output** (torch.FloatTensor): A output sequence of encoder. `FloatTensor` of size
``(seq_length, dimension)``
**max_lengths** (int): Max decoding time step
Returns:
**decode_output** (torch.LongTensor): Result of model predictions
"""
token_list = list()
hidden_states = None
token = torch.LongTensor([[self.decoder.sos_id]])
if torch.cuda.is_available():
token = token.cuda()
for i in range(max_lengths):
decoder_output, hidden = self.decoder(token, hidden_states=hidden_states)
output = self.joint(encoder_output[i].view(-1), decoder_output.view(-1))
prediction_token = output.topk(1)[1]
token = prediction_token.unsqueeze(1) # (1, 1)
prediction_token = int(prediction_token.item())
token_list.append(prediction_token)
return torch.LongTensor(token_list)
is bug or not for this code:
decoder_output, hidden = self.decoder(token, hidden_states=hidden_states)
hidden is nerver used. hidden should be hidden_states?
์ฝ๋๋ฅผ ๋๋ ค๋ณด์ง๋ ์์์ง๋ง ๋ด๋ถ์ 1d conv layer๊ฐ depthwise ๋ ์ด์ด์ธ ๊ฒ์ผ๋ก ์๋๋ฐ,
๊ทธ๋ ๋ค๋ฉด in_channels = out_channnels = groups๊ฐ ๋์ผ ํ๋ ๊ฒ ์๋๊ฐ์?
I check the dimension of the files. I find that we don't need to write this code 'output = inputs.transpose(1, 2)' in the file 'audio_encoder.py'. Instead,we write 'output = inputs'. Iโm not sure if my idea is right. I want to know what the author's idea is.
Great work @upskyy Very easy to follow implementation. A couple of questions and I appreciate the clarification:
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.