ThilinaRajapakse/simpletransformers

Wraps HuggingFace Transformers with task-specific model classes that standardize the train/eval/predict workflow across NLP and multi-modal applications. Built-in integrations with Weights & Biases enable experiment tracking, while support for any HuggingFace pretrained model (BERT, RoBERTa, T5, etc.) provides flexibility without lock-in. Dense retrieval, conversational AI, and encoder fine-tuning extend beyond typical classification pipelines.

4,234 stars and 52,813 monthly downloads. Used by 4 other packages. No commits in the last 6 months. Available on PyPI.