yesj1234/xlsr_mid1_ko-ja

en-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: 0.3889
• Cer: 0.1082

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: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1500
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
1.4503	1.22	2000	1.0610	0.2687
1.0239	2.45	4000	0.6962	0.1904
0.8977	3.67	6000	0.5945	0.1687
0.804	4.9	8000	0.5328	0.1492
0.698	6.12	10000	0.5014	0.1365
0.6426	7.35	12000	0.4715	0.1322
0.61	8.57	14000	0.4530	0.1258
0.5709	9.79	16000	0.4300	0.1201
0.5235	11.02	18000	0.4168	0.1166
0.4778	12.24	20000	0.4057	0.1129
0.4571	13.47	22000	0.3945	0.1100
0.4388	14.69	24000	0.3891	0.1081

Framework versions

• Transformers 4.34.0
• Pytorch 2.1.0+cu121
• Datasets 2.14.5
• Tokenizers 0.14.1