msalhab96 / speeq Goto Github PK
View Code? Open in Web Editor NEWA framework for automatic speech recognition
Home Page: https://speeq.readthedocs.io/
License: MIT License
A framework for automatic speech recognition
Home Page: https://speeq.readthedocs.io/
License: MIT License
Adding script that does data re-structuring and formatting to be used and trained in the framework
No response
No response
No response
No response
No response
No response
Currently the data is not sorted neither get shuffled during training
No response
Enhance CSVDataset class to have ',' as default sep.
Before the enhancment:
from speeq.data.loaders import CSVDataset
file = CSVDataset(data_path="test.csv", sep=',')
After the enhancement:
from speeq.data.loaders import CSVDataset
file = CSVDataset(data_path="test.csv")
No response
The current RNN based decoder with attention rely on the the last hidden state when it comes to the first prediction or injecting the SOS token, this should be resolved
N/A
0.0.0
No response
No response
Currently, the only tokenizer supported is a character based tokenizer, it would be good to add white space/word tokenizer
No response
currently, there's no way to integrate or use any augmentation implemented with the data pipeline
N/A
0.0.0
No response
No response
An object from ctc.py::CTCModel can be instantiated, but the forward function raises an Error because self.encoder is not implemented.
Should be modified to private.
No response
No response
No response
No response
Line 359 in 6ef6da6
SpeechTransformer uses Transformer decoder, which does not match the decoder used in paper.
SpeechTransformerEncoder implementation missing LayerNorm
Currently when the trainer args get created the gradient clipping params get ignored!
No response
No response
No response
No response
Currently we support only Transducer builder and CTC builder only
No response
Currently, a pre-trained model get loaded silently, it would be better to add a log message indicates that the model is loadded
No response
Currently, the optimizer parameters get loaded if a pre-trained model passed, it would be better if we Add an option to reset the optimizer learning rate when loading it
No response
When loading a pre-trained model to continue the training from the latest checkpoint the min_loss in the trainer gets assigned to inf
it has to get the latest value reached from the history!
No response
No response
No response
No response
Adding gradient accumulation across training steps in the trainers
No response
the parameters text_key and speech_key in the SpeechTextDataset are not getting initialized from the data config object and they always take the default value
No response
No response
No response
No response
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.