yesj1234/xlsr_5p_ko-zh

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: 0.5172
• Cer: 0.1076

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
6.742	2.6	1500	0.9038	0.2330
0.9228	5.2	3000	0.6193	0.1656
0.6805	7.8	4500	0.5522	0.1481
0.5577	10.4	6000	0.5136	0.1349
0.4797	13.0	7500	0.5074	0.1312
0.4161	15.6	9000	0.4959	0.1243
0.3701	18.21	10500	0.4948	0.1224
0.3307	20.81	12000	0.4881	0.1199
0.2946	23.41	13500	0.4970	0.1179
0.2636	26.01	15000	0.4950	0.1145
0.2367	28.61	16500	0.4905	0.1119
0.2157	31.21	18000	0.5066	0.1110
0.1979	33.81	19500	0.5133	0.1103
0.1808	36.41	21000	0.5160	0.1091
0.17	39.01	22500	0.5129	0.1074

Framework versions

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