profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/lorenlugosch/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
Loren Lugosch lorenlugosch Ph.D. student researching machine learning at McGill University and Mila.

speechbrain/speechbrain 3027

A PyTorch-based Speech Toolkit

speechbrain/speechbrain.github.io 284

The SpeechBrain project aims to build a novel speech toolkit fully based on PyTorch. With SpeechBrain users can easily create speech processing systems, ranging from speech recognition (both HMM/DNN and end-to-end), speaker recognition, speech enhancement, speech separation, multi-microphone speech processing, and many others.

lorenlugosch/end-to-end-SLU 193

PyTorch code for end-to-end spoken language understanding (SLU) with ASR-based transfer learning

lorenlugosch/pytorch_HMM 70

HMMs in PyTorch

lorenlugosch/neural-min-sum-decoding 23

Neural min-sum decoding in TensorFlow

lorenlugosch/transducer-tutorial 9

Example code for a neural transducer model.

lorenlugosch/infer_missing_vowels 5

This repository contains code for training a simple model to infer the original sentence when vowels are removed.

lorenlugosch/open-speech-corpora 2

A list of accessible speech corpora for ASR, TTS, and other Speech Technologies

lebrice/VHDL-CPU 1

5-stage pipeline, 32-bit MIPS CPU written in VHDL for the Computer Organization & Architecture (ECSE425) Course at McGill University

push eventlorenlugosch/speechbrain

Nauman Dawalatabad

commit sha 4866138cdf72c769d2a5a045ffbe5332d9768eb3

Running some experiments

view details

Nauman Dawalatabad

commit sha 9092934009ca58bd67d4c057d4cc11a0b7beb264

Adding beamforming

view details

Nauman Dawalatabad

commit sha 8c8daad61c0739ba2c0901111955be5e34184b01

Reverting to "Adding beamforming" as ecapa yaml got lost somewhere This reverts commit 9092934009ca58bd67d4c057d4cc11a0b7beb264.

view details

Nauman Dawalatabad

commit sha 93efac3e1dcc97641c0c8532433d754bd6073891

Adding beamformer with yaml files

view details

Nauman Dawalatabad

commit sha 4a046c060c5eda58925b63b155bf5d25aef1de7f

removing prefix Mix from data_preparation

view details

Nauman Dawalatabad

commit sha aedfd8d22b803009ed02de7a96d44f6e407534bb

running exps for Distant mic data

view details

Nauman Dawalatabad

commit sha 65f62bdeeb09cf305380956fa91947dac62f537b

Completed exps with distant mics with ecapa. Saving hash for future reference

view details

Nauman Dawalatabad

commit sha 7f78f494876527a8c4c5eb27e99a5bb7edfcf93f

exps comparing BeamformIt from Kaldi

view details

Nauman Dawalatabad

commit sha 9ef5a30cb63ea0b285c6a3885769b05bffa7e3f4

Added backend kmeans support

view details

Nauman Dawalatabad

commit sha 7dd9fa7fc0838be839fed41ea1dad719ba967712

Exp for different backends

view details

Nauman Dawalatabad

commit sha ea9f39d2ba0754e01f85f3363c7c6ea119589671

exp with backend

view details

Nauman Dawalatabad

commit sha 9d7c1fe310d3d2bc9d2faf179261aa6de1f77784

ahc addition

view details

Nauman Dawalatabad

commit sha 06b944c5b0e5b4d1c5e578b816774af84e0e7232

Updated logic inside main for accomodating ahc.

view details

Nauman Dawalatabad

commit sha 5ce3729471d8216501746f13c3d61b16131af077

minor

view details

Nauman Dawalatabad

commit sha 32822062fec796d386c139c2f76a001309a4c974

switching to euclidean linkage

view details

Nauman Dawalatabad

commit sha f4ee87a89ca196f982c64333f843ba7a276a6901

minor

view details

Nauman Dawalatabad

commit sha 44f57854d02401b8c825b751b9f60df00496e1ac

experiment_beam.py code refactoring

view details

Nauman Dawalatabad

commit sha 1d2f0dac4aba7e021dd9ae4ccbefa5e32d63a0df

fixing bug in dev tuning

view details

Nauman Dawalatabad

commit sha 03ce4f779b2c9de2dca194291feeb85e5b486d10

Fixing issue #578. added check_dir()

view details

Nauman Dawalatabad

commit sha c559153eb858a31987229e35487ea85567abce08

adding new yaml for ecapa

view details

push time in a month

startedsyhw/wer_are_we

started time in 2 months

push eventlorenlugosch/speechbrain

Jianyuan Zhong

commit sha 01868aa9a950e688f6d29f4747ec76d1b1b9217e

checkpoint avg

view details

Parcollet Titouan

commit sha 729ae44365627b898f03b0d3f9f8a0e6f8441396

Update index.rst

view details

Parcollet Titouan

commit sha 75465b15dc0fb6f6f01eac2c46120d381c5fc78d

Update README.md

view details

Parcollet Titouan

commit sha 59e6e783e0c6eb2f34fa30855199965cee25d8aa

Update README.md

view details

Parcollet Titouan

commit sha 1ec7a166f28f001c06832a8666e7068c7496e557

Update README.md

view details

Parcollet Titouan

commit sha 98f02d959350aac4ce6353e9aceeb3b3b01b8808

Update installation.md

view details

Haoxian WANG

commit sha cd9e13ef348aae30622b3b98fdbec4a950c4b553

Adding feature allowing torchaudio to use soundfile on windows and sox_io on mac/linux.

view details

Haoxian WANG

commit sha e15f5c6ecd2f72adacceb4f005feb742b9d978a3

fix trailing whitespaces.

view details

Haoxian WANG

commit sha 658a2a03726d325c2ab3a706f2b290745e1240ae

formatted code

view details

Haoxian WANG

commit sha f09474942d8e0fe265f7af8f29fd12cf8d5b28cc

formatted code

view details

Haoxian WANG

commit sha 20d6282afc50fa86d187cbde5cba7a73008c1d06

formatted code

view details

Haoxian WANG

commit sha ee762ee4273d4dfebfbfc2469aeebc15a0875fcc

formatted code for arguments

view details

Haoxian WANG

commit sha fb0fda54f87e0ad6684c4affc0f850f3d55d878d

added torch_audio_backend.py in utils for retrieving the system information.

view details

Haoxian WANG

commit sha ee7a527750c24a6dcde0df9c7292c38a2a6715f2

formatted torch_audio_backend.py

view details

Työ

commit sha 034b306d414ef603de48dbf19c05ad5b3510409e

Move to same source and conf directory for RTD

view details

Työ

commit sha 7d17da94f8f08650c6ce5fbd173b6b92006afc6b

Remove tools from toctree - not generated

view details

JianyuanZhong

commit sha 7014130e3383cfa30805ec59053a522825bbff60

Update README.md

view details

Jianyuan Zhong

commit sha 3b9564cf26d5cf413dbd26ebd0b8a424670fd3f0

Merge branch 'update-transformer' into update-read-me

view details

Parcollet Titouan

commit sha 34bcf9d0783cf72a952674032834383194018b7b

Merge pull request #564 from speechbrain/update-read-me Update README.md

view details

Jianyuan Zhong

commit sha f2ce190c0117740b0b2ded96e149a482925cc254

added checkpint avg with auto merge averaged checkpoints

view details

push time in 2 months

push eventlorenlugosch/speechbrain

popcornell

commit sha fb82e77f856368b400b0968cb9c6c53f1fa65d81

added MHA with relative pos encodings as in Conformer and TransformerXL + tests. Need Docs

view details

cem

commit sha f4b46802a3e60c169e302f5f13957604364c6156

Merge branch 'develop' of github.com:speechbrain/speechbrain into RelPosMHA

view details

cem

commit sha daa8489e5259eab6b35c065dcd4c98931d247aa1

added the option to use RelativeMultiHeadAttention inside transformer

view details

cem

commit sha 309686f36654e74c02856af9fa0817c246e965bb

adding sepformer-relativeembs.yaml

view details

cem

commit sha 5c96e02d0bb807f29710ed3181ddd06611bd0fce

Merge branch 'develop' of github.com:speechbrain/speechbrain into RelPosMHA

view details

cem

commit sha f1411540650a76535bd2a4210bde23130cb0aa26

fixed the bug in transformer encoder layer

view details

Jianyuan Zhong

commit sha d27363a45025c4c9a67cda0d0fded833399e90aa

fix shape; masking

view details

Jianyuan Zhong

commit sha 67da6a8d369417f2ff95e628c4ecd27774d2eb54

added checkpoint avg

view details

Jianyuan Zhong

commit sha ccf3ad938ff615e1cea9feae00ee842bee979e36

fix tests

view details

Mirco Ravanelli

commit sha 9a6fb7e8609c53934b1042ea5fad051be2ab41b0

Merge pull request #546 from speechbrain/conformer-asr Conformer asr

view details

Jianyuan Zhong

commit sha cd998c9a18f3d8fb96c6a30885793eb633d496ed

fix pos encoding

view details

JianyuanZhong

commit sha 057f85f2a37f6e4f64e214391d245ec70a61e7d0

Update README.md

view details

JianyuanZhong

commit sha 478b00ecf1f6756e8782f3ddcce8252281aedce0

Update transformer.yaml

view details

popcornell

commit sha 829a983814cacba2a5e454c7e2b926037b72a940

added mask also on positions

view details

popcornell

commit sha bbe18b59200e020d4e9649f21d812bcaa5251d20

keeping transformer architecture for now

view details

popcornell

commit sha 0d35d44d59264a8d98b9b72d26f17b696c069ce0

added transformerXL style pos encoding

view details

popcornell

commit sha f90259a78d81cc10442447e8392154ee77a62d64

apply linter

view details

popcornell

commit sha d53445b21c9af2e4ab2f953535724457558b9569

two times faster than other implementations

view details

popcornell

commit sha 2feae9cb8cb9b7ad0501183be8e344a928351750

fixed key_padding_mask

view details

popcornell

commit sha d81b50b410e89a1a8421eb53ae11af7c4bdafff8

added pos_embs argument also to regular MHA as we will need it in future

view details

push time in 2 months

issue commentspeechbrain/speechbrain

how to support fp16 in transducer_loss.py

Ah, we also have a LibriSpeech transducer recipe, but it doesn't work very well: https://github.com/speechbrain/speechbrain/tree/develop/recipes/LibriSpeech/ASR/transducer

jinggaizi

comment created time in 3 months

issue commentspeechbrain/speechbrain

how to support fp16 in transducer_loss.py

PER for a transducer recipe is reported for TIMIT here: https://github.com/speechbrain/speechbrain/tree/develop/recipes/TIMIT/ASR/transducer

Don't know if we have any other transducer recipes

jinggaizi

comment created time in 3 months

pull request commentspeechbrain/speechbrain

tokenizer from SB paper (58 tokens instead of 51 for complete coverage)

@mravanelli , do you know how to change the link from our results drive to something SB can download (for the new tokenizer)? It's https://drive.google.com/u/0/uc?id=1muXFLLqHrKWWXAjF_R0WuukZro9KWgt3&export=download

lorenlugosch

comment created time in 3 months

create barnchlorenlugosch/speechbrain

branch : fix-slurp-tokenizer

created branch time in 3 months

push eventlorenlugosch/speechbrain

Titouan

commit sha 15608603ba15598c6c731718a852673f4793ce43

updating models

view details

Parcollet Titouan

commit sha ecfd2c5a90de14eae2e409ea39e62e92bb87fb64

Update conftest.py

view details

Parcollet Titouan

commit sha ffeb18033e916b2b17d1322d543cbc53fc8d44ef

Update conftest.py

view details

DavidRP1

commit sha 6742f64d913a1dfe7b00156247ce4ecece8617a4

Include GSC recipe

view details

cem

commit sha 59bc3bf412dc5c1c2e9baf687ede623cc1c4c588

initial implementation of dynamic whamr reverberation

view details

cem

commit sha c37726984795f12d8a435c3a8e4f3b913261f61e

updated so that we use SB reverberate function

view details

cem

commit sha aba562193fe251e55c29991701c168764d20127e

implemented a speechbrain wrapper for convtranspose1d

view details

ravanelm

commit sha da426b855d09af3bc18f8b5b1d621c630590deff

recipe normalization

view details

ravanelm

commit sha a02999ba9295c6a0fccbaaf26034de2900b290e6

fixed style

view details

ravanelm

commit sha b7ff9dc4102dffebd75330fd91d375943fcc51dc

folder change

view details

mirco

commit sha c49a501ba92b2a6054cc47547d26dd6861c5732f

add an example with conv1d + convtransp1d

view details

ravanelm

commit sha bd2e64d6f6b598dc8dd61c71202c45204ac38433

fixed prep + update README

view details

cem

commit sha 3fbf6b7a32003a6a2ea8f9583382c3298cd8b1de

added the output size option to convtranspose1d

view details

cem

commit sha 5763cd40a8195712d839d869d69e351d1ae1d05c

added valid/same options to convtranspose1d

view details

cem

commit sha aa8f3b38dd9490264a042265accf3478449ebb15

added name

view details

Peter Plantinga

commit sha e5bbf231779d5f48c87aa81983f9b0f54b9776df

Updates to sequential class to clarify

view details

Loren Lugosch

commit sha 930c64fd6dd94e3e7aae8509759af4243fe2621b

Remove comma preprocessing, add script for running multiple experiments, load checkpoint with best valid accuracy, add testing on all real splits

view details

Loren Lugosch

commit sha 62ad99f97261e9ff924355131730999ee0702b00

test black

view details

Loren Lugosch

commit sha 30a36f523be0ee0f066d514a58a9f438434cf7ab

add comment

view details

lorenlugosch

commit sha e89d944f55980bf4e01fe36b6d0839fc1a4e12c9

black

view details

push time in 3 months