bert-base-finnish-cased-v1 for QA

This is the bert-base-finnish-cased-v1 model, fine-tuned using the Finnish SQuAD2.0 dataset. It's been trained on question-answer pairs, including unanswerable questions, for the task of Question Answering.

Overview

Language model: bert-base-finnish-v1

Language: Finnish

Downstream-task: Extractive QA

Training data: squad_v2_fi

Eval data: squad_v2_fi

Hyperparameters

batch_size = 12
n_epochs = 2
base_LM_model = "bert-base-finnish-cased-v1"
max_seq_len = 384
learning_rate = 3e-5
doc_stride=128

Performance

"exact": 68.2424,
"f1": 73.6607,
"HasAns_exact": 58.3290,
"HasAns_f1": 69.3629,
"HasAns_total": 5817,
"NoAns_exact": 77.8073,
"NoAns_f1": 77.8073,
"NoAns_total": 6029,
"samples": 12138,
"total": 11846

Comparison with other Finnish models