review_classification_bert_base_jp_v3_ratio1_10_2label_add_dropout-epoch30_v4

This model is a fine-tuned version of cl-tohoku/bert-base-japanese-v3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.0927
• Accuracy: 0.9003

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: 1e-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
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	286	0.2417	0.9075
0.2411	2.0	572	0.2688	0.9131
0.2411	3.0	858	0.3000	0.9126
0.1763	4.0	1144	0.3900	0.8768
0.1763	5.0	1430	0.5771	0.8952
0.0758	6.0	1716	0.7230	0.8921
0.0273	7.0	2002	0.7799	0.8799
0.0273	8.0	2288	0.8761	0.8911
0.0046	9.0	2574	1.0195	0.8742
0.0046	10.0	2860	0.9275	0.8952
0.0078	11.0	3146	0.9748	0.8906
0.0078	12.0	3432	0.9619	0.8988
0.0057	13.0	3718	1.0354	0.8865
0.0038	14.0	4004	1.1006	0.8747
0.0038	15.0	4290	0.9827	0.8998
0.0016	16.0	4576	1.0556	0.8891
0.0016	17.0	4862	1.1509	0.8793
0.0027	18.0	5148	1.2724	0.8630
0.0027	19.0	5434	1.0569	0.8931
0.0041	20.0	5720	1.1100	0.8911
0.0022	21.0	6006	1.0499	0.9024
0.0022	22.0	6292	1.0498	0.9054
0.0017	23.0	6578	1.1034	0.8937
0.0017	24.0	6864	1.0687	0.9024
0.0005	25.0	7150	1.0940	0.8972
0.0005	26.0	7436	1.0940	0.8978
0.0003	27.0	7722	1.0938	0.8993
0.0007	28.0	8008	1.1729	0.8885
0.0007	29.0	8294	1.0835	0.9024
0.0014	30.0	8580	1.0927	0.9003

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.15.0
• Tokenizers 0.15.0