yesj1234/xlsr_5p_ko-en

ko-xlsr

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the ./SAMPLE_SPEECH.PY - NA dataset. It achieves the following results on the evaluation set:

• Loss: 0.4651
• Cer: 0.0828

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: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
6.3673	3.17	1500	0.6104	0.1606
0.656	6.33	3000	0.4318	0.1129
0.4729	9.5	4500	0.4010	0.1028
0.3789	12.66	6000	0.3867	0.0977
0.3166	15.83	7500	0.3857	0.0936
0.267	18.99	9000	0.3891	0.0912
0.2286	22.16	10500	0.4074	0.0910
0.1967	25.32	12000	0.4079	0.0878
0.1712	28.49	13500	0.4289	0.0865
0.1493	31.65	15000	0.4456	0.0850
0.1333	34.82	16500	0.4573	0.0843
0.1191	37.98	18000	0.4633	0.0833

Framework versions

• Transformers 4.34.0
• Pytorch 2.0.1+cu117
• Datasets 2.14.5
• Tokenizers 0.14.0