bluebert_pubmed_mimic_uncased_squadv2

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

bluebert_pubmed_mimic_uncased_squadv2

This model is a fine-tuned version of bionlp/bluebert_pubmed_mimic_uncased_L-12_H-768_A-12 on the squad_v2 dataset.

Intended uses & limitations

This is the first model on huggingface that combines MIMIC data with squadv2 for question answering purposes.

Training and evaluation data

Takes the pretrained model bionlp/bluebert_pubmed_mimic_uncased_L-12_H-768_A-12 and fine-tunes and evaluates on squad_v2 data.

Training procedure

Tuning script used (.bat file):

@echo off

set BASE_MODEL=bionlp/bluebert_pubmed_mimic_uncased_L-12_H-768_A-12
set OUTPUT_DIR=U:\Documents\Breast_Non_Synoptic\results\pretrained\bluebert_pubmed_mimic_uncased_squadv2\

python run_qa.py ^
  --model_name_or_path  %BASE_MODEL% ^
  --dataset_name squad_v2 ^
  --do_train ^
  --do_eval ^
  --version_2_with_negative ^
  --per_device_train_batch_size 16 ^
  --learning_rate 2e-5 ^
  --num_train_epochs 3 ^
  --max_seq_length 480 ^
  --doc_stride 64 ^
  --weight_decay 0.01 ^
  --output_dir %OUTPUT_DIR%

You may need to adapt this script for non-Windows operating systems.

The run_qa.py example script can be found here.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.1
• Datasets 2.14.4
• Tokenizers 0.13.2