xlm-roberta-base-finetuned-JaQuAD

This model is a fine-tuned version of xlm-roberta-base on the JaQuAD dataset. It achieves the following results on the evaluation set:

Loss: 0.7495

Model description

More information needed

Intended uses

import torch
from transformers import AutoModelForQuestionAnswering, AutoTokenizer

model_name = "thkkvui/xlm-roberta-base-finetuned-JaQuAD"
model = (AutoModelForQuestionAnswering.from_pretrained(model_name))
tokenizer = AutoTokenizer.from_pretrained(model_name)

text = "2015年9月1日、私は横浜へ車で出かけました。映画を観た後に中華街まで電車で行き、昼ご飯は重慶飯店で中華フルコースを食べました。"
questions= ["どこへ出かけた？", "電車に乗る前は何をしていた？", "重慶飯店で何を食べた？", "いつ横浜に出かけた？"]

for question in questions:
    
    inputs = tokenizer.encode_plus(question, text, add_special_tokens=True, return_tensors="pt")
    
    with torch.no_grad():
        output = model(**inputs)

    answer_start = torch.argmax(output.start_logits)
    answer_end = torch.argmax(output.end_logits)

    answer_tokens = inputs.input_ids[0, answer_start : answer_end + 1]
    answer = tokenizer.decode(answer_tokens)

    print(f"質問: {question} -> 回答: {answer}")

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 6e-05
train_batch_size: 16
eval_batch_size: 16
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 50
num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss
0.8661	1.0	1985	0.8036
0.5348	2.0	3970	0.7495

Framework versions

Transformers 4.30.2
Pytorch 2.0.1
Datasets 2.13.1
Tokenizers 0.13.3