Metadata-Version: 2.1
Name: transformers-collection
Version: 0.2.0
Summary: A collection of transformer models built  using huggingface for various tasks.
Home-page: https://github.com/aadhithya/transformers-collection
License: MIT
Keywords: transformers,huggingface,NLP
Author: Aadhithya Sankar
Author-email: aadhithya.s@outlook.com
Requires-Python: >=3.9,<3.11
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Dist: datasets (>=2.8.0,<3.0.0)
Requires-Dist: evaluate (>=0.4.0,<0.5.0)
Requires-Dist: loguru (>=0.6.0,<0.7.0)
Requires-Dist: munch (>=2.5.0,<3.0.0)
Requires-Dist: numpy (>=1.24.1,<2.0.0)
Requires-Dist: pytorch-lightning (>=1.9.0,<2.0.0)
Requires-Dist: scikit-learn (>=1.2.1,<2.0.0)
Requires-Dist: scipy (>=1.10.0,<2.0.0)
Requires-Dist: tensorboard (>=2.11.2,<3.0.0)
Requires-Dist: torch (>1.10,<2.0)
Requires-Dist: transformers (>=4.26.0,<5.0.0)
Requires-Dist: typer (>=0.7.0,<0.8.0)
Project-URL: Repository, https://github.com/aadhithya/transformers-collection
Description-Content-Type: text/markdown

# transformers-collection
- A collection of transformer models built  using huggingface for various tasks. Training done using pytorch lightning.
- Datasets, models and tokenizers from hugging face.
- **Goal**: Get familiar with huggingface and pytorch lightning ecosystems.

## Get started
### Train Models using the library
- To train models, install using pip: `pip install transformers_collection`
- check installation: `transformers-collection version`

### Clone project and modify code
To play around with the code clone the repo:
- `git clone git@github.com:aadhithya/transformers-collection.git`
- Install poetry: `pip install poetry`
- Intsall dependencies: `poetry install`

**Note:** `poetry install` will create a new venv.
**Note**: `poetry/pip install` installs CPU version of pytorch if not available, please make sure to install CUDA version if needed.


## Train a model
- Create the yaml config file for the model (see configs/sentiment-clf.yml for example).
- train model using: `transformers-collection train /path/to/config.yml`

- For a list of supported models, see section Supported Models.



## Supported Models / Task
The following models are planned:
| Model                            |                      Dataset                       |  Status   | Checkpoint |
| :------------------------------- | :------------------------------------------------: | :-------: | ---------: |
| Sentiment/Emotion Classification | [emotion](https://huggingface.co/datasets/emotion) |     ✅     |        TBD |
| Text Summarization               |                                                    | 🗓️ Planned |        TBD |

