Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/calclavia/tal-asrd
/ types & classes
Types & classes
35 in github.com/calclavia/tal-asrd
⨍
Functions
314
◇
Types & classes
35
↓ 6 callers
Class
ASRAlignedDataset
Audio -> Text dataset. Provides well aligned audio utterances. Two special tokens are used. Text format: Begin episode
tal/asr/data/aligned.py:76
↓ 5 callers
Class
ASRAlignedCollater
Collates data
tal/asr/data/aligned.py:240
↓ 4 callers
Class
System
tal/asr/system.py:26
↓ 3 callers
Class
Tokenizer
tal/asr/tokenizers/sentencepiece.py:17
↓ 3 callers
Class
WandbLoggerWrapper
tal/asr/logger.py:5
↓ 2 callers
Class
BeamState
Structure that contains necessary states for beam search.
tal/diarization/uisrnn/uisrnn.py:42
↓ 2 callers
Class
LabelSmoothLoss
tal/asr/util.py:19
↓ 2 callers
Class
Lamb
r""" https://github.com/cybertronai/pytorch-lamb Implements Lamb algorithm. It has been proposed in `Large Batch Optimization for Deep Lea
tal/optimizers.py:302
↓ 2 callers
Class
LogMelSpec
Performs log melspec transformation on raw audio Adapted from https://github.com/zcaceres/spec_augment/blob/master/SpecAugment.ipynb P
tal/asr/models.py:15
↓ 2 callers
Class
ModRZTXDecoderLayer
Modified version of decoder layer that caches the attention weights for analysis
tal/asr/models.py:488
↓ 2 callers
Class
SDUtteranceCollater
Collates data
tal/asr/data/baseline_speaker.py:214
↓ 2 callers
Class
SDUtteranceDataset
Audio -> Text dataset. Provides well aligned audio utterances. Two special tokens are used. Text format: Begin episode <EOS>
tal/asr/data/baseline_speaker.py:75
↓ 2 callers
Class
System
tal/baseline/speaker_system.py:25
↓ 2 callers
Class
TDS
Acoustic model
tal/asr/models.py:349
↓ 1 callers
Class
ASRModel
tal/asr/models.py:56
↓ 1 callers
Class
ASRSegmentDataset
Audio -> Text dataset. During training we obtain a random segment of audio and ask the model to predict all utterances within that c
tal/asr/data/segment.py:56
↓ 1 callers
Class
CoreRNN
The core Recurent Neural Network used by UIS-RNN.
tal/diarization/uisrnn/uisrnn.py:20
↓ 1 callers
Class
DefaultArgs
tal/asr/transcribe.py:213
↓ 1 callers
Class
Frame
Represents a "frame" of audio data.
tal/asr/speech_detect.py:36
↓ 1 callers
Class
Frame
Represents a "frame" of audio data.
tal/vad/webrtcvad.py:49
↓ 1 callers
Class
PositionalEncoding
r"""Inject some information about the relative or absolute position of the tokens in the sequence. The positional encodings have the same dim
tal/modules.py:24
↓ 1 callers
Class
SDModel
tal/asr/models.py:400
↓ 1 callers
Class
SuppressPrint
tal/__init__.py:43
↓ 1 callers
Class
System
tal/asr/train_embed.py:21
↓ 1 callers
Class
TDSBlock
tal/asr/models.py:298
Class
ASRSegmentCollater
Collates data
tal/asr/data/segment.py:304
Class
Adafactor
Fairseq implementation Implements Adafactor algorithm. https://github.com/pytorch/fairseq/blob/master/fairseq/optim/adafactor.py This
tal/optimizers.py:84
Class
AudioCollator
Collates data
tal/asr/data/audio.py:107
Class
Logger
A class for printing logging information to screen.
tal/diarization/uisrnn/utils.py:15
Class
Lookahead
tal/optimizers.py:234
Class
RAdam
tal/optimizers.py:9
Class
RandomSegmentDataset
A dataset that produces random audio segments from a directory
tal/asr/data/audio.py:29
Class
Tokenizer
tal/asr/tokenizers/transformers.py:11
Class
UISRNN
Unbounded Interleaved-State Recurrent Neural Networks.
tal/diarization/uisrnn/uisrnn.py:67
Class
_Tokenizer
tal/asr/tokenizers/__init__.py:7