bert-base-japanese-ghost_rate-weighted

This model is a fine-tuned version of cl-tohoku/bert-base-japanese on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.7735
• Accuracy: 0.4195
• F1: 0.4186

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
No log	0.9977	220	1.5524	0.3651	0.3644
No log	2.0	441	1.5018	0.3781	0.3761
1.5438	2.9977	661	1.5367	0.3934	0.3832
1.5438	4.0	882	1.5724	0.4019	0.4042
1.1647	4.9977	1102	1.6488	0.4093	0.4115
1.1647	6.0	1323	1.7250	0.4138	0.4176
0.8864	6.9977	1543	1.7735	0.4195	0.4186
0.8864	8.0	1764	1.8372	0.4138	0.4161
0.8864	8.9977	1984	1.8975	0.4150	0.4133
0.6978	9.9773	2200	1.8925	0.4127	0.4146

Framework versions

• Transformers 4.40.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1