This is the starting code for the default final project for the Stanford CS 224N class. You can find the handout here
In this project, you will implement some important components of the BERT model to better understanding its architecture. You will then use the embeddings produced by your BERT model on three downstream tasks: sentiment classification, paraphrase detection and semantic similarity.
After finishing the BERT implementation, you will have a simple model that simultaneously performs the three tasks. You will then implement extensions to improve on top of this baseline.
- Follow
setup.shto properly setup a conda environment and install dependencies. - There is a detailed description of the code structure in STRUCTURE.md, including a description of which parts you will need to implement.
- You are only allowed to use libraries that are installed by
setup.sh, external libraries that give you other pre-trained models or embeddings are not allowed (e.g.,transformers).
Please refer to the handout for a through description of the project and its parts.
The BERT implementation part of the project was adapted from the "minbert" assignment developed at Carnegie Mellon University's CS11-711 Advanced NLP, created by Shuyan Zhou, Zhengbao Jiang, Ritam Dutt, Brendon Boldt, Aditya Veerubhotla, and Graham Neubig.
Parts of the code are from the transformers library (Apache License 2.0).