This model, referred to as LastBERT, is a lightweight yet potent BERT-based model designed for natural language processing (NLP) applications. It was created through knowledge distillation from a larger BERT model, resulting in a significant reduction in parameters—from 110 million in BERT-base-uncased to just 29 million in LastBERT, making it approximately 73.64% smaller. Despite its reduced size, LastBERT maintains robust performance across various tasks, such as paraphrase identification, sentiment analysis, and grammatical acceptability, as demonstrated on the General Language Understanding Evaluation (GLUE) benchmark.

Downloads last month
19
Safetensors
Model size
29.8M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.
The model cannot be deployed to the HF Inference API: The model has no library tag.