Model Description

BioMobileBERT is the result of training the MobileBERT-uncased model in a continual learning scenario for 200k training steps using a total batch size of 192 on the PubMed dataset.

Initialisation

We initialise our model with the pre-trained checkpoints of the MobileBERT-uncased model available on Huggingface.

Architecture

MobileBERT uses a 128-dimensional embedding layer followed by 1D convolutions to up-project its output to the desired hidden dimension expected by the transformer blocks. For each of these blocks, MobileBERT uses linear down-projection at the beginning of the transformer block and up-projection at its end, followed by a residual connection originating from the input of the block before down-projection. Because of these linear projections, MobileBERT can reduce the hidden size and hence the computational cost of multi-head attention and feed-forward blocks. This model additionally incorporates up to four feed-forward blocks in order to enhance its representation learning capabilities. Thanks to the strategically placed linear projections, a 24-layer MobileBERT (which is used in this work) has around 25M parameters.

Citation

If you use this model, please consider citing the following paper:

@misc{https://doi.org/10.48550/arxiv.2209.03182,
  doi = {10.48550/ARXIV.2209.03182},
  url = {https://arxiv.org/abs/2209.03182},
  author = {Rohanian, Omid and Nouriborji, Mohammadmahdi and Kouchaki, Samaneh and Clifton, David A.},
  keywords = {Computation and Language (cs.CL), Machine Learning (cs.LG), FOS: Computer and information sciences, FOS: Computer and information sciences, 68T50},
  title = {On the Effectiveness of Compact Biomedical Transformers},
  publisher = {arXiv},
  year = {2022}, 
  copyright = {arXiv.org perpetual, non-exclusive license}
}