22221fwfwfw

This model is a fine-tuned version of thisisHJLee/wav2vec2-large-xls-r-300m-korean-g on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.0469
• Cer: 0.1656

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: 0.0003
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 400
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.2455	1.62	500	1.0127	0.2441
0.3772	3.25	1000	1.0766	0.2435
0.3073	4.87	1500	0.9578	0.2234
0.2446	6.49	2000	0.9579	0.2175
0.2069	8.12	2500	1.1064	0.2111
0.1769	9.74	3000	1.1212	0.2124
0.1561	11.36	3500	1.1560	0.2137
0.1368	12.99	4000	1.1001	0.2097
0.1226	14.61	4500	0.9872	0.1920
0.1069	16.23	5000	0.9882	0.1909
0.098	17.86	5500	1.0663	0.1899
0.0846	19.48	6000	1.0553	0.1896
0.0754	21.1	6500	1.0569	0.1847
0.0657	22.73	7000	1.0256	0.1741
0.0606	24.35	7500	1.0654	0.1745
0.0501	25.97	8000	1.0614	0.1779
0.0438	27.6	8500	1.0316	0.1690
0.039	29.22	9000	1.0469	0.1656

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu118
• Datasets 2.13.0
• Tokenizers 0.13.3