Bert Tiny Finetuned Mrpc

🚀 BERT-tiny model finetuned with M-FAC

This is a BERT-tiny model finetuned on the MRPC dataset using the state-of-the-art second-order optimizer M-FAC, offering enhanced performance for text classification tasks.

🚀 Quick Start

This model is finetuned on the MRPC dataset with the state-of-the-art second-order optimizer M-FAC. For more details on M-FAC, check the NeurIPS 2021 paper: https://arxiv.org/pdf/2107.03356.pdf.

✨ Features

• Advanced Optimization: Utilizes the M-FAC optimizer for finetuning, which is a state-of-the-art second-order optimization method.
• Fair Comparison Setup: Finetuned in the same framework as the default Adam baseline for fair performance comparison.

📦 Installation

No specific installation steps are provided in the original README. If you want to use this model, you may need to refer to the related repositories:

• Hugging Face Transformers: https://github.com/huggingface/transformers
• M-FAC Code: https://github.com/IST-DASLab/M-FAC

💻 Usage Examples

Finetuning setup

For fair comparison against the default Adam baseline, we finetune the model in the same framework as described here https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification and just swap the Adam optimizer with M-FAC.

Hyperparameters used by the M-FAC optimizer:

learning rate = 1e-4
number of gradients = 512
dampening = 1e-6

Results Reproduction

The results can be reproduced by adding the M-FAC optimizer code in https://github.com/huggingface/transformers/blob/master/examples/pytorch/text-classification/run_glue.py and running the following bash script:

CUDA_VISIBLE_DEVICES=0 python run_glue.py \
  --seed 42 \
  --model_name_or_path prajjwal1/bert-tiny \
  --task_name mrpc \
  --do_train \
  --do_eval \
  --max_seq_length 128 \
  --per_device_train_batch_size 32 \
  --learning_rate 1e-4 \
  --num_train_epochs 5 \
  --output_dir out_dir/ \
  --optim MFAC \
  --optim_args '{"lr": 1e-4, "num_grads": 512, "damp": 1e-6}'

📚 Documentation

Results

We share the best model out of 5 runs with the following score on the MRPC validation set:

f1 = 83.12
accuracy = 73.52

Mean and standard deviation for 5 runs on the MRPC validation set:

We believe these results could be improved with modest tuning of hyperparameters: per_device_train_batch_size, learning_rate, num_train_epochs, num_grads and damp. For the sake of fair comparison and a robust default setup, we use the same hyperparameters across all models (bert-tiny, bert-mini) and all datasets (SQuAD version 2 and GLUE).

📄 License

No license information is provided in the original README.

📚 Citation

@article{frantar2021m,
  title={M-FAC: Efficient Matrix-Free Approximations of Second-Order Information},
  author={Frantar, Elias and Kurtic, Eldar and Alistarh, Dan},
  journal={Advances in Neural Information Processing Systems},
  volume={35},
  year={2021}
}