SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Wav2Vec2 meets phonemes (#14353)

Browse Source 
* up

* add tokenizer

* improve more

* finish tokenizer

* finish

* adapt speech recognition script

* adapt convert

* more fixes

* more fixes

* update phonemizer wav2vec2

* better naming

* fix more tests

* more fixes swedish

* correct tests

* finish

* improve script

* remove file

* up

* lets get those 100 model architectures until the end of the month

* make fix-copies

* correct more

* correct script

* more fixes

* more fixes

* add to docs

* Apply suggestions from code review

Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* replace assert

* fix copies

* fix docs

* new try docs

* boom boom

* update

* add phonemizer to audio tests

* make fix-copies

* up

* upload models

* some changes

* Update tests/test_tokenization_wav2vec2_phoneme.py

Co-authored-by: Anton Lozhkov <aglozhkov@gmail.com>

* more fixes

* remove @

Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Anton Lozhkov <aglozhkov@gmail.com>
This commit is contained in:
Patrick von Platen
2021-12-17 19:56:44 +01:00
committed by GitHub
parent 77d6c826d8
commit c4a96cecbc
26 changed files with 1296 additions and 151 deletions
Download Patch File Download Diff File Expand all files Collapse all files

56
docs/source/model_doc/wav2vec2_phoneme.mdx Normal file
View File

@@ -0,0 +1,56 @@
<!--Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
-->
# Wav2Vec2Phoneme
## Overview
The Wav2Vec2Phoneme model was proposed in [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition (Xu et al.,
2021](https://arxiv.org/abs/2109.11680) by Qiantong Xu, Alexei Baevski, Michael Auli.
The abstract from the paper is the following:
*Recent progress in self-training, self-supervised pretraining and unsupervised learning enabled well performing speech
recognition systems without any labeled data. However, in many cases there is labeled data available for related
languages which is not utilized by these methods. This paper extends previous work on zero-shot cross-lingual transfer
learning by fine-tuning a multilingually pretrained wav2vec 2.0 model to transcribe unseen languages. This is done by
mapping phonemes of the training languages to the target language using articulatory features. Experiments show that
this simple method significantly outperforms prior work which introduced task-specific architectures and used only part
of a monolingually pretrained model.*
Tips:
- Wav2Vec2Phoneme uses the exact same architecture as Wav2Vec2
- Wav2Vec2Phoneme is a speech model that accepts a float array corresponding to the raw waveform of the speech signal.
- Wav2Vec2Phoneme model was trained using connectionist temporal classification (CTC) so the model output has to be
decoded using [`Wav2Vec2PhonemeCTCTokenizer`].
- Wav2Vec2Phoneme can be fine-tuned on multiple language at once and decode unseen languages in a single forward pass
to a sequence of phonemes
- By default the model outputs a sequence of phonemes. In order to transform the phonemes to a sequence of words one
should make use of a dictionary and language model.
Relevant checkpoints can be found under https://huggingface.co/models?other=phoneme-recognition.
This model was contributed by [patrickvonplaten](https://huggingface.co/patrickvonplaten)
The original code can be found [here](https://github.com/pytorch/fairseq/tree/master/fairseq/models/wav2vec).
Wav2Vec2Phoneme's architecture is based on the Wav2Vec2 model, so one can refer to [`Wav2Vec2`]'s documentation page except for the tokenizer.
## Wav2Vec2PhonemeCTCTokenizer
[[autodoc]] Wav2Vec2PhonemeCTCTokenizer
- __call__
- batch_decode
- decode
- phonemize

47
docs/source/model_doc/xls_r.rst Normal file
View File

@@ -0,0 +1,47 @@
..
Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
XLS-R
-----------------------------------------------------------------------------------------------------------------------
Overview
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The XLS-R model was proposed in `XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale
<https://arxiv.org/abs/2111.09296>`__ by Arun Babu, Changhan Wang, Andros Tjandra, Kushal Lakhotia, Qiantong Xu, Naman
Goyal, Kritika Singh, Patrick von Platen, Yatharth Saraf, Juan Pino, Alexei Baevski, Alexis Conneau, Michael Auli.
The abstract from the paper is the following:
*This paper presents XLS-R, a large-scale model for cross-lingual speech representation learning based on wav2vec 2.0.
We train models with up to 2B parameters on nearly half a million hours of publicly available speech audio in 128
languages, an order of magnitude more public data than the largest known prior work. Our evaluation covers a wide range
of tasks, domains, data regimes and languages, both high and low-resource. On the CoVoST-2 speech translation
benchmark, we improve the previous state of the art by an average of 7.4 BLEU over 21 translation directions into
English. For speech recognition, XLS-R improves over the best known prior work on BABEL, MLS, CommonVoice as well as
VoxPopuli, lowering error rates by 14-34% relative on average. XLS-R also sets a new state of the art on VoxLingua107
language identification. Moreover, we show that with sufficient model size, cross-lingual pretraining can outperform
English-only pretraining when translating English speech into other languages, a setting which favors monolingual
pretraining. We hope XLS-R can help to improve speech processing tasks for many more languages of the world.*
Tips:
- XLS-R is a speech model that accepts a float array corresponding to the raw waveform of the speech signal.
- XLS-R model was trained using connectionist temporal classification (CTC) so the model output has to be decoded using
:class:`~transformers.Wav2Vec2CTCTokenizer`.
Relevant checkpoints can be found under https://huggingface.co/models?other=xls_r.
XLS-R's architecture is based on the Wav2Vec2 model, so one can refer to :doc:`Wav2Vec2's documentation page
<wav2vec2>`.
The original code can be found `here <https://github.com/pytorch/fairseq/tree/master/fairseq/models/wav2vec>`__.