đ Whisper Telugu Large-v2
This model is a fine - tuned version of openai/whisper-large-v2 on Telugu data from multiple publicly available ASR corpuses. It was fine - tuned as part of the Whisper fine - tuning sprint.
đ Quick Start
This model is a fine - tuned version of openai/whisper-large-v2 on the Telugu data sourced from multiple publicly available Automatic Speech Recognition (ASR) corpuses. It was developed as part of the Whisper fine - tuning sprint.
â ī¸ Important Note
The code used to train this model is available for re - use in the whisper-finetune repository.
⨠Features
- Fine - tuned on multiple Telugu ASR corpuses.
- Code for training, evaluation, and faster inference is available in the whisper-finetune repository.
đĻ Installation
No specific installation steps are provided in the original document.
đģ Usage Examples
Basic Usage
In order to infer a single audio file using this model, the following code snippet can be used:
>>> import torch
>>> from transformers import pipeline
>>>
>>> audio = "/path/to/audio.format"
>>> device = "cuda:0" if torch.cuda.is_available() else "cpu"
>>> transcribe = pipeline(task="automatic-speech-recognition", model="vasista22/whisper-telugu-large-v2", chunk_length_s=30, device=device)
>>> transcribe.model.config.forced_decoder_ids = transcribe.tokenizer.get_decoder_prompt_ids(language="te", task="transcribe")
>>> print('Transcription: ', transcribe(audio)["text"])
Advanced Usage
For faster inference of whisper models, the whisper-jax library can be used. Please follow the necessary installation steps as mentioned here, before using the following code snippet:
>>> import jax.numpy as jnp
>>> from whisper_jax import FlaxWhisperForConditionalGeneration, FlaxWhisperPipline
>>>
>>> audio = "/path/to/audio.format"
>>> transcribe = FlaxWhisperPipline("vasista22/whisper-telugu-large-v2", batch_size=16)
>>> transcribe.model.config.forced_decoder_ids = transcribe.tokenizer.get_decoder_prompt_ids(language="te", task="transcribe")
>>> print('Transcription: ', transcribe(audio)["text"])
đ Documentation
In order to evaluate this model on an entire dataset, the evaluation codes available in the whisper-finetune repository can be used. The same repository also provides the scripts for faster inference using whisper - jax.
đ§ Technical Details
Training and Evaluation Data
Property |
Details |
Training Data |
CSTD IIIT - H ASR Corpus, [ULCA ASR Corpus](https://github.com/Open - Speech - EkStep/ULCA - asr - dataset - corpus#telugu - labelled - total - duration - is - 102593 - hours), Shrutilipi ASR Corpus, [Microsoft Speech Corpus (Indian Languages)](https://msropendata.com/datasets/7230b4b1 - 912d - 400e - be58 - f84e0512985e), Google/Fleurs Train+Dev set, Babel ASR Corpus |
Evaluation Data |
[Microsoft Speech Corpus (Indian Languages) Test Set](https://msropendata.com/datasets/7230b4b1 - 912d - 400e - be58 - f84e0512985e), Google/Fleurs Test Set, OpenSLR, Babel Test Set |
Training Hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.75e - 05
- train_batch_size: 8
- eval_batch_size: 32
- seed: 22
- optimizer: adamw_bnb_8bit
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 22000
- training_steps: 75000
- mixed_precision_training: True
đ License
This model is licensed under the Apache - 2.0 license.
Acknowledgement
This work was done at Speech Lab, IIT Madras. The compute resources for this work were funded by "Bhashini: National Language translation Mission" project of the Ministry of Electronics and Information Technology (MeitY), Government of India.