by Zhengbao Jiang, Shuyan Zhou, and Ritam Dutt

This is an exercise in developing a minimalist version of BERT, part of Carnegie Mellon University's CS11-747: Neural Networks for NLP.

In this assignment, you will implement some important components of the BERT model to better understanding its architecture. You will then perform sentence classification on sst dataset and cfimdb dataset with the BERT model.

• Follow setup.sh to properly setup the 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 torch, no other external libraries are allowed (e.g., transformers).
• We will run your code with the following commands, so make sure that whatever your best results are reproducible using these commands (where you replace ANDREWID with your andrew ID):

mkdir -p ANDREWID

python3 classifier.py --option [pretrain/finetune] --epochs NUM_EPOCHS --lr LR --train data/sst-train.txt --dev data/sst-dev.txt --test data/sst-test.txt

Pretraining for SST: Dev Accuracy : 0.400 Test Accuracy : 0.414

Mean reference accuracies over 10 random seeds with their standard deviation shown in brackets.

Finetuning for SST : Dev Accuracy : 0.520 (0.006) Test Accuracy : 0.525 (0.007)

The submission file should be a zip file with the following structure (assuming the andrew id is ANDREWID):

ANDREWID/
├── base_bert.py
├── bert.py
├── classifier.py
├── config.py
├── optimizer.py
├── sanity_check.py
├── tokenizer.py
├── utils.py
├── README.md
├── structure.md
├── sanity_check.data
├── sst-dev-output.txt 
├── sst-test-output.txt 
├── cfimdb-dev-output.txt 
├── cfimdb-test-output.txt 
└── setup.py

• A+: You additionally implement something else on top of the requirements for A, and achieve significant accuracy improvements:
  • perform continued pre-training using the MLM objective to do domain adaptation
  • try alternative fine-tuning algorithms
  • add other model components on top of the model
• A: You implement all the missing pieces and the original classifier.py with --option finetune code that achieves comparable accuracy to our reference implementation
• A-: You implement all the missing pieces and the original classifier.py with --option pretrain code that achieves comparable accuracy to our reference implementation
• B+: All missing pieces are implemented and pass tests in sanity_check.py, but accuracy is not comparable to the reference.
• B or below: Some parts of the missing pieces are not implemented.

Parts of the code are from the transformers library (Apache License 2.0).