hubert-rinnna-jp-jdrtsp-fw07sp-14

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

• Loss: 0.1685
• Wer: 0.2927
• Cer: 0.1710

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: 16
• 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
• lr_scheduler_warmup_steps: 1000
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
4.4197	1.0	404	4.0767	0.9928	0.9960
2.8984	2.0	808	2.7950	0.9928	0.9960
2.1179	3.0	1212	1.9178	0.9928	0.9960
1.4282	4.0	1616	1.0854	0.6262	0.4791
1.0793	5.0	2020	0.7672	0.4996	0.2944
0.9064	6.0	2424	0.6212	0.4573	0.2737
0.8366	7.0	2828	0.5247	0.4132	0.2450
0.7425	8.0	3232	0.4502	0.3786	0.2257
0.7017	9.0	3636	0.3912	0.3509	0.2082
0.6275	10.0	4040	0.3407	0.3328	0.1979
0.5853	11.0	4444	0.3045	0.3226	0.1920
0.5551	12.0	4848	0.2657	0.3139	0.1865
0.5105	13.0	5252	0.2455	0.3086	0.1827
0.5073	14.0	5656	0.2389	0.3092	0.1832
0.4722	15.0	6060	0.2170	0.3030	0.1781
0.481	16.0	6464	0.2089	0.3023	0.1783
0.4738	17.0	6868	0.2002	0.3004	0.1763
0.4518	18.0	7272	0.1990	0.3006	0.1765
0.4402	19.0	7676	0.1900	0.2999	0.1764
0.4387	20.0	8080	0.1826	0.2970	0.1740
0.4212	21.0	8484	0.1767	0.2955	0.1733
0.3893	22.0	8888	0.1707	0.2937	0.1719
0.4055	23.0	9292	0.1704	0.2943	0.1723
0.394	24.0	9696	0.1684	0.2934	0.1716
0.3997	25.0	10100	0.1685	0.2927	0.1710

Framework versions

• Transformers 4.34.0.dev0
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.13.3