A study to benchmark whisper based ASRs in Malayalam
Project description
malayalam_asr_benchmarking
The work is still in progress. I have now done some benchmarking for Common Voice 11 Malayalam dataset. The benchmarking results has been uploaded to hugging face as a dataset. At the moment I am working on benchmarking Malayalam Speech Corpus dataset as well. The benchmarking results once completed will be uploaded to huggingface datasets in the same manner.
Install
pip install malayalam_asr_benchmarking
Or locally
pip install -e .
Setting up your development environment
I am developing this project with nbdev. Please take some time reading up on nbdev … how it works, directives, etc… by checking out the walk-thrus and tutorials on the nbdev website
Step 1: Install Quarto:
nbdev_install_quarto
Other options are mentioned in getting started to quarto
Step 2: Install hooks
nbdev_install_hooks
Step 3: Install our library
pip install -e '.[dev]'
How to use
from malayalam_asr_benchmarking.commonvoice import evaluate_whisper_model_common_voice
evaluate_whisper_model_common_voice("parambharat/whisper-tiny-ml")
Found cached dataset common_voice_11_0 (/home/.cache/huggingface/datasets/mozilla-foundation___common_voice_11_0/ml/11.0.0/2c65b95d99ca879b1b1074ea197b65e0497848fd697fdb0582e0f6b75b6f4da0)
Loading cached processed dataset at /home/.cache/huggingface/datasets/mozilla-foundation___common_voice_11_0/ml/11.0.0/2c65b95d99ca879b1b1074ea197b65e0497848fd697fdb0582e0f6b75b6f4da0/cache-374585c2877047e3.arrow
Loading cached processed dataset at /home/.cache/huggingface/datasets/mozilla-foundation___common_voice_11_0/ml/11.0.0/2c65b95d99ca879b1b1074ea197b65e0497848fd697fdb0582e0f6b75b6f4da0/cache-22670505c562e0d4.arrow
/opt/conda/lib/python3.8/site-packages/transformers/generation_utils.py:1359: UserWarning: Neither `max_length` nor `max_new_tokens` has been set, `max_length` will default to 448 (`self.config.max_length`). Controlling `max_length` via the config is deprecated and `max_length` will be removed from the config in v5 of Transformers -- we recommend using `max_new_tokens` to control the maximum length of the generation.
warnings.warn(
Total time taken: 133.23447608947754
The WER of model: 38.31
The CER of model: 21.93
The model size is: 37.76M
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for malayalam_asr_benchmarking-0.0.2.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | 109106e4e58dfc0f312548758f7b8b3c2658fd9238245d2c187b2935f9c91d47 |
|
MD5 | f9dd2ce683bdc2c962417ba8adf6ccfb |
|
BLAKE2b-256 | 8282e453c255028ca97fc55b00e4abd861958de7f8a8bf8422ea04091627581a |
Hashes for malayalam_asr_benchmarking-0.0.2-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 94bc218f4d480824483819eded03f7ca07605c7d38daad8a7277477e6d4d9941 |
|
MD5 | fab6ca8ff9ab8963ddc5278b20fc33b2 |
|
BLAKE2b-256 | 88d31d0f0a86a7f0fcf0c0b74fa167728bb8345760ac40320d3ddc0ad4e88dc0 |