🚀 BERT-mini model finetuned with M-FAC
This project presents a BERT-mini model finetuned with the state-of-the-art second-order optimizer M-FAC on the MNLI dataset, offering enhanced performance.
🚀 Quick Start
This model is finetuned on the MNLI dataset using the state-of-the-art second-order optimizer M-FAC. For more details about M-FAC, refer to the NeurIPS 2021 paper: https://arxiv.org/pdf/2107.03356.pdf.
✨ Features
- The model is finetuned using M-FAC, a second - order optimizer, which may lead to better performance compared to traditional optimizers.
- The finetuning setup is designed for a fair comparison with the default Adam baseline.
📦 Installation
There is no specific installation step provided in the original README. So, this section is skipped.
💻 Usage Examples
Basic Usage
To reproduce the results, you need to add the M-FAC optimizer code in https://github.com/huggingface/transformers/blob/master/examples/pytorch/text-classification/run_glue.py and run the following bash script:
CUDA_VISIBLE_DEVICES=0 python run_glue.py \
--seed 8276 \
--model_name_or_path prajjwal1/bert-mini \
--task_name mnli \
--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": 1024, "damp": 1e-6}'
📚 Documentation
Finetuning setup
For a 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.
The hyperparameters used by the M-FAC optimizer are:
learning rate = 1e-4
number of gradients = 1024
dampening = 1e-6
Results
We share the best model out of 5 runs with the following score on the MNLI validation set:
matched_accuracy = 75.13
mismatched_accuracy = 75.93
The mean and standard deviation for 5 runs on the MNLI validation set are as follows:
Property |
Matched Accuracy |
Mismatched Accuracy |
Adam |
73.30 ± 0.20 |
74.85 ± 0.09 |
M-FAC |
74.59 ± 0.41 |
75.95 ± 0.14 |
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).
Our code for M-FAC can be found here: https://github.com/IST-DASLab/M-FAC.
A step-by-step tutorial on how to integrate and use M-FAC with any repository can be found here: https://github.com/IST-DASLab/M-FAC/tree/master/tutorials.
📄 License
There is no license information provided in the original README. So, this section is skipped.
🔧 Technical Details
There is no specific technical details section with more than 50 - word description in the original README. So, this section is skipped.
BibTeX entry and citation info
@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}
}