All Projects → nttcslab-sp → torchain

nttcslab-sp / torchain

Licence: other
WIP: pytorch FFI wrapper for Kaldi chain loss (a.k.a. Lattice Free MMI)

Programming Languages

python
139335 projects - #7 most used programming language
C++
36643 projects - #6 most used programming language
shell
77523 projects
Makefile
30231 projects
c
50402 projects - #5 most used programming language

Projects that are alternatives of or similar to torchain

Zamia Speech
Open tools and data for cloudless automatic speech recognition
Stars: ✭ 374 (+1770%)
Mutual labels:  kaldi, asr
rustfst
Rust re-implementation of OpenFST - library for constructing, combining, optimizing, and searching weighted finite-state transducers (FSTs). A Python binding is also available.
Stars: ✭ 104 (+420%)
Mutual labels:  kaldi, asr
Eesen
The official repository of the Eesen project
Stars: ✭ 738 (+3590%)
Mutual labels:  kaldi, asr
Vosk Android Demo
Offline speech recognition for Android with Vosk library.
Stars: ✭ 271 (+1255%)
Mutual labels:  kaldi, asr
Py Kaldi Asr
Some simple wrappers around kaldi-asr intended to make using kaldi's (online) decoders as convenient as possible.
Stars: ✭ 156 (+680%)
Mutual labels:  kaldi, asr
Vosk Server
WebSocket, gRPC and WebRTC speech recognition server based on Vosk and Kaldi libraries
Stars: ✭ 277 (+1285%)
Mutual labels:  kaldi, asr
Espresso
Espresso: A Fast End-to-End Neural Speech Recognition Toolkit
Stars: ✭ 808 (+3940%)
Mutual labels:  kaldi, asr
opensnips
Open source projects related to Snips https://snips.ai/.
Stars: ✭ 50 (+150%)
Mutual labels:  kaldi, asr
Speech To Text Russian
Проект для распознавания речи на русском языке на основе pykaldi.
Stars: ✭ 151 (+655%)
Mutual labels:  kaldi, asr
Pytorch Asr
ASR with PyTorch
Stars: ✭ 124 (+520%)
Mutual labels:  kaldi, asr
Docker Kaldi Gstreamer Server
Dockerfile for kaldi-gstreamer-server.
Stars: ✭ 266 (+1230%)
Mutual labels:  kaldi, asr
Zeroth
Kaldi-based Korean ASR (한국어 음성인식) open-source project
Stars: ✭ 248 (+1140%)
Mutual labels:  kaldi, asr
kaldi-alligner
scripts to align a given wave to its transcription using trained models by Kaldi
Stars: ✭ 24 (+20%)
Mutual labels:  kaldi, asr
Asr theory
语音识别理论,论文和PPT
Stars: ✭ 344 (+1620%)
Mutual labels:  kaldi, asr
vosk-model-ru-adaptation
No description or website provided.
Stars: ✭ 19 (-5%)
Mutual labels:  kaldi, asr
Pykaldi
A Python wrapper for Kaldi
Stars: ✭ 756 (+3680%)
Mutual labels:  kaldi, asr
kaldi-long-audio-alignment
Long audio alignment using Kaldi
Stars: ✭ 21 (+5%)
Mutual labels:  kaldi, asr
Vosk Api
Offline speech recognition API for Android, iOS, Raspberry Pi and servers with Python, Java, C# and Node
Stars: ✭ 1,357 (+6685%)
Mutual labels:  kaldi, asr
Pytorch Kaldi
pytorch-kaldi is a project for developing state-of-the-art DNN/RNN hybrid speech recognition systems. The DNN part is managed by pytorch, while feature extraction, label computation, and decoding are performed with the kaldi toolkit.
Stars: ✭ 2,097 (+10385%)
Mutual labels:  kaldi, asr
asr24
24-hour Automatic Speech Recognition
Stars: ✭ 27 (+35%)
Mutual labels:  kaldi, asr

PyTorch binding of Kaldi Lattice-Free MMI Loss

Currently this project is developed with pytroch 0.4.0 and kaldi 22fbdd (CHiME5 baseline). We will support pytorch 1.0.0 and latest kaldi soon.

$ make test KALDI_ROOT=<your kaldi-22fbdd>

then you will see stdout like this

0 ChainResults(loss=1523440.000000, objf=-65544.968750, l2_term=-1523429.125000, weight=6016.000000)
1 ChainResults(loss=90851.000000, objf=-11141.706055, l2_term=-90848.437500, weight=4352.000000)
2 ChainResults(loss=1633.794800, objf=-5968.860352, l2_term=-1632.802612, weight=6016.000000)
3 ChainResults(loss=824.511902, objf=-6661.337402, l2_term=-823.548157, weight=6912.000000) 
4 ChainResults(loss=411.817657, objf=-4280.249023, l2_term=-410.834137, weight=4352.000000) 
5 ChainResults(loss=408.518799, objf=-5801.333496, l2_term=-407.554474, weight=6016.000000) 
6 ChainResults(loss=233.465958, objf=-4343.767090, l2_term=-232.467850, weight=4352.000000) 
7 ChainResults(loss=243.598633, objf=-4366.199707, l2_term=-242.595367, weight=4352.000000) 
8 ChainResults(loss=387.405365, objf=-5805.361816, l2_term=-386.440369, weight=6016.000000) 
9 ChainResults(loss=253.922623, objf=-4391.189453, l2_term=-252.913620, weight=4352.000000) 
10 ChainResults(loss=389.562714, objf=-6743.458496, l2_term=-388.587097, weight=6912.000000)
11 ChainResults(loss=271.677155, objf=-5866.956543, l2_term=-270.701935, weight=6016.000000)
12 ChainResults(loss=203.969925, objf=-4334.875000, l2_term=-202.973862, weight=4352.000000)
13 ChainResults(loss=297.204376, objf=-5742.427734, l2_term=-296.249847, weight=6016.000000)
14 ChainResults(loss=190.700623, objf=-4397.112305, l2_term=-189.690262, weight=4352.000000)
15 ChainResults(loss=160.930191, objf=-4469.153809, l2_term=-159.903275, weight=4352.000000)
16 ChainResults(loss=196.294601, objf=-5793.267578, l2_term=-195.331619, weight=6016.000000)
17 ChainResults(loss=243.681747, objf=-6467.249512, l2_term=-242.746094, weight=6912.000000)
18 ChainResults(loss=158.063736, objf=-4443.177734, l2_term=-157.042786, weight=4352.000000)
19 ChainResults(loss=181.009201, objf=-5824.614258, l2_term=-180.041016, weight=6016.000000)
20 ChainResults(loss=113.364677, objf=-4284.602051, l2_term=-112.380165, weight=4352.000000) 

TODO

known issues

  • torchain.io.open_example return nullptr when batchsize is changed
Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].