output

This model is a fine-tuned version of /home/son/Work/wav2vec2-xls-r-300m/facebook/wav2vec2-xls-r-300m on the KRESNIK/ZEROTH_KOREAN - CLEAN dataset. It achieves the following results on the evaluation set:

• Loss: 2.1666
• Wer: 0.9737
• Cer: 0.5039

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: 7.5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: 2000
• num_epochs: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
19.558	1.44	500	19.4094	1.0	1.0
4.7968	2.87	1000	4.7828	1.0	1.0
4.5125	4.31	1500	4.4959	0.9991	0.9540
4.2202	5.75	2000	4.2905	0.9923	0.8520
3.7774	7.18	2500	3.2846	1.0356	0.6652
3.1418	8.62	3000	2.3624	0.9882	0.5429

Framework versions

• Transformers 4.24.0
• Pytorch 1.13.1
• Datasets 2.6.1
• Tokenizers 0.11.0