Comments (6)
It looks like pinyin with tone (0, 1, 2, 3, 4) added, one way to represent the pronunication for Chinese characters.
from nemo.
So may I use same (0, 1, 2, 3, 4) for my none English language (Armenian)?
Currently my lables look like
labels: [' ', '!', '"', "'", '(', ')', ',', '-', '.', ':', ';', '?', '՜', '՞', 'Ա', 'Բ', 'Գ', 'Դ', 'Ե', 'Զ', 'Է', 'Ը', 'Թ', 'Ժ', 'Ի', 'Լ', 'Խ', 'Ծ', 'Կ', 'Հ', 'Ձ', 'Ղ', 'Ճ', 'Մ', 'Յ', 'Ն', 'Շ', 'Ո', 'Չ', 'Պ', 'Ջ', 'Ռ', 'Ս', 'Վ', 'Տ', 'Ր', 'Ց', 'Ւ', 'Փ', 'Ք', 'Օ', 'Ֆ', 'ա', 'բ', 'գ', 'դ', 'ե', 'զ', 'է', 'ը', 'թ', 'ժ', 'ի', 'լ', 'խ', 'ծ', 'կ', 'հ', 'ձ', 'ղ', 'ճ', 'մ', 'յ', 'ն', 'շ', 'ո', 'չ', 'պ', 'ջ', 'ռ', 'ս', 'վ', 'տ', 'ր', 'ց', 'ւ', 'փ', 'ք', 'օ', 'ֆ', 'և']
from nemo.
I think if your labels can cover the all the characters which appear in the training dataset, then it is ok.. You don't need to add the tone symbols (which seems only appear in Chinese)
Reference:
https://en.wikipedia.org/wiki/Wikipedia:Language_recognition_charthttps://en.wikipedia.org/wiki/Wikipedia:Language_recognition_chart
from nemo.
Thank you very much!
You support is very quick and effective.
from nemo.
@ican24 Hi, The input to the Tacotron2 should in some way contain the pronunciation information in your language. For example, we use 'pinyin' to represent the pronunciation of a Chinese character. For a Chinese character '好', its 'pinyin' is 'hao3'. For a Chinese phrase '你好', its 'pinyin' is 'ni3 hao3'.
There are tens of thousands of different Chinese characters. If we simply use the Chinese characters as input. There will be tens of thousands of labels and it will be so hard for the model to learn. Instead, we use the pinyin as input so the input labels are just 26 English characters + 5 tones (0-4) + some symbols. Which should make Tacotron2 easier to learn.
So In your case. Since there are only 38 characters in Armenian. You can try directly using the characters of Armenian as input(label).
from nemo.
Dear Muyang Du,
Thank you for your detailed explication and confirmation!
Any detail is so important before launching long term processes like tacotron2.py and waveglow.
from nemo.
Related Issues (20)
- Tips for training Ecapa-TDNN speaker recognition model as fast as possible?
- Support LLAMA 3.1 in NeMo HOT 1
- AttributeError: 'dict' object has no attribute '_get_node_flag' HOT 1
- LazyNeMoIterator improperly duration handling for non-zero offset in nemo manifest file HOT 8
- STT_EN_FASTCONFORMER_TRANSDUCER_XLARG - Throws error for Tensor + List operation in confidence calculation HOT 1
- Getting TypeError: EncDecCTCModel.transcribe() got an unexpected keyword argument 'logprobs'
- finetune nvidia/parakeet-tdt-1.1b results out of memeory even if with lower batch size. HOT 1
- TypeError: TransformerBlockSubmodules.__init__() got an unexpected keyword argument 'layer_norm' HOT 1
- Add CI Tests for Canary/AEDMultitask "lang_field" HOT 1
- Fine-Tuning Fast Conformer CTC on Other Language HOT 2
- How to Enable Spec Augmentation During Testing or Transcription HOT 1
- pytorch_lightning ResultCollection register_key is deprecated in 2.4.0 HOT 1
- [question] What text preprocessing have you used for ASR models stt-eu, stt-de, stt-es, stt-fr. Is it only nemo text norm? HOT 1
- TypeError: AggregateTokenizer.token_to_id() missing 1 required positional argument: 'lang_id' HOT 2
- Eval_beamsearch_ngram_ctc throws got an unexpected keyword argument 'logprobs' HOT 3
- FSDP is not integrated across code base HOT 2
- " ValueError: max() arg is an empty sequence " while converting mamba 2 hybrid checkpoint to nemo HOT 2
- Out of RAM using 24.07 container HOT 4
- training resume is not working in 24.07 container HOT 2
- slurm Multi-machine and multi-GPU training
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from nemo.