https://github.com/luluw8071/automatic-speech-recognition-with-pytorch

End-to-End Automatic Speech Recognition on PyTorch with CTC Decoder and Ken LM
https://github.com/luluw8071/automatic-speech-recognition-with-pytorch

asr-model cnn-lstm-models ctc-decode cuda-support deep-neural-networks kenlm python pytorch pytorch-lightning

Last synced: 2 days ago
JSON representation

End-to-End Automatic Speech Recognition on PyTorch with CTC Decoder and Ken LM

• Host: GitHub
• URL: https://github.com/luluw8071/automatic-speech-recognition-with-pytorch
• Owner: LuluW8071
• License: gpl-3.0
• Created: 2023-07-30T16:18:56.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2024-10-21T15:25:31.000Z (24 days ago)
• Last Synced: 2024-10-22T02:15:28.038Z (23 days ago)
• Topics: asr-model, cnn-lstm-models, ctc-decode, cuda-support, deep-neural-networks, kenlm, python, pytorch, pytorch-lightning
• Language: Python
• Homepage:
• Size: 155 KB
• Stars: 4
• Watchers: 2
• Forks: 1
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# End-to-End Automatic Speech Recognition



![Code in Progress](https://img.shields.io/badge/status-completed-green.svg) ![License](https://img.shields.io/github/license/LuluW8071/Automatic-Speech-Recognition-with-PyTorch) ![Open Issues](https://img.shields.io/github/issues/LuluW8071/Automatic-Speech-Recognition-with-PyTorch) ![Closed Issues](https://img.shields.io/github/issues-closed/LuluW8071/Automatic-Speech-Recognition-with-PyTorch) ![Open PRs](https://img.shields.io/github/issues-pr/LuluW8071/Automatic-Speech-Recognition-with-PyTorch) ![Repo Size](https://img.shields.io/github/repo-size/LuluW8071/Deep-Speech-2) ![Last Commit](https://img.shields.io/github/last-commit/LuluW8071/Automatic-Speech-Recognition-with-PyTorch)



This project implements a small scale speech recognition system utilizing a Residual Convolutional Neural Network (CNN) - BiGRU Acoustic Model, a Connectionist Temporal Classification (CTC) Decoder, and a KENLM Language Model for enhanced accuracy.

## Model Architecture

## Installation

1. Clone the repository:

   ```bash

   git clone --recursive https://github.com/LuluW8071/Automatic-Speech-Recognition-with-PyTorch.git

   ```

2. Install **[Pytorch](https://pytorch.org/)** and required dependencies under virtual environment:

   ```bash

   pip install -r requirements.txt

   ```

   Ensure you have `PyTorch` and `Lightning AI` installed.

## Train Model

>[!IMPORTANT]

> Before training make sure you have placed __comet ml api key__ and __project name__ in the environment variable file `.env`.

```bash

py train.py

```

Customize the pytorch training parameters by passing arguments in `train.py` to suit your needs:

Refer to the provided table to change hyperparameters and train configurations.

| Args                   | Description                                                           | Default Value      |

|------------------------|-----------------------------------------------------------------------|--------------------|

| `-g, --gpus`           | Number of GPUs per node                                               | 1  |

| `-g, --num_workers`           | Number of CPU workers                                               | 8  |

| `-db, --dist_backend`           | Distributed backend to use for training                             | ddp_find_unused_parameters_true  |

| `--epochs`             | Number of total epochs to run                                         | 50                 |

| `--batch_size`         | Size of the batch                                                     | 32                |

| `-lr, --learning_rate`      | Learning rate                                                         | 1e-5  (0.00001)      |

| `--checkpoint_path` | Checkpoint path to resume training from                                 | None |

| `--precision`        | Precision of the training                                              | 16-mixed |

```bash

py train.py 

-g 4                   # Number of GPUs per node for parallel gpu training

-w 8                   # Number of CPU workers for parallel data loading

--epochs 10            # Number of total epochs to run

--batch_size 64        # Size of the batch

-lr 2e-5               # Learning rate

--precision 16-mixed   # Precision of the training

```

>[!NOTE]

>To __resume training__ from a saved checkpoint, use:

```bash

py train.py --checkpoint_path path_to_checkpoint.ckpt

```

## Additional Resources

For pre-trained models and other resources, refer to the provided links.

[Click here to download pre trained model](https://mega.nz/folder/Lnxj3YCJ#Na6Nc1m4nz6jiSWTatfKJQ)

---

This comprehensive guide should help you navigate through setting up and using the Speech Recognition system effectively. If you encounter any issues or have questions, feel free to reach out!