metadata

license: cc-by-sa-4.0
base_model: retrieva-jp/t5-base-long
tags:
  - generated_from_trainer
  - summarization
datasets:
  - csebuetnlp/xlsum
metrics:
  - rouge
model-index:
  - name: t5-base-xlsum-ja
    results:
      - task:
          name: Sequence-to-sequence Language Modeling
          type: text2text-generation
        dataset:
          name: csebuetnlp/xlsum
          type: xlsum
          config: japanese
          split: test
          args: japanese
        metrics:
          - name: Rouge1
            type: rouge
            value: 0.3648008957585529
          - name: Rouge2
            type: rouge
            value: 0.16411161798042992
language:
  - ja
library_name: transformers

t5-base-xlsum-ja

This model is a fine-tuned version of retrieva-jp/t5-base-long on the xlsum dataset. It achieves the following results on the evaluation set:

Loss: 2.6563
Rouge1: 0.3648
Rouge2: 0.1641
Rougel: 0.2965
Rougelsum: 0.3132

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0001
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 16
total_train_batch_size: 128
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.01
num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
4.9166	1.8	100	3.4095	0.3569	0.1509	0.2416	0.3209
4.1162	3.61	200	3.0980	0.3262	0.1354	0.2557	0.2805
3.8578	5.41	300	2.8853	0.3428	0.1445	0.2628	0.2881
3.7309	7.22	400	2.7714	0.3621	0.1615	0.2951	0.3151
3.6716	9.02	500	2.7042	0.3727	0.1668	0.2982	0.3225
3.6393	10.82	600	2.6666	0.3676	0.1592	0.2987	0.3206
3.6291	12.63	700	2.6587	0.3654	0.1576	0.2955	0.3108
3.6224	14.43	800	2.6563	0.3648	0.1641	0.2965	0.3132

Framework versions

Transformers 4.34.0
Pytorch 2.0.0+cu118
Datasets 2.14.5
Tokenizers 0.14.0