yesj1234/xlsr_jako_exp1

jako-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: 0.9486
• Cer: 0.2606

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: 2
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
3.5667	1.14	1000	2.2323	0.5188
1.5569	2.28	2000	1.3106	0.3527
1.2238	3.43	3000	1.1109	0.3099
1.0593	4.57	4000	1.0390	0.2891
0.9658	5.71	5000	0.9731	0.2918
0.8796	6.85	6000	0.9479	0.2696
0.8022	8.0	7000	0.9331	0.2710
0.7392	9.14	8000	0.9252	0.2746
0.6694	10.28	9000	0.9318	0.2590
0.5977	11.42	10000	0.9349	0.2674
0.5484	12.56	11000	0.9409	0.2555
0.5154	13.71	12000	0.9510	0.2719
0.4767	14.85	13000	0.9556	0.2624
0.4536	15.99	14000	0.9850	0.2684
0.4195	17.13	15000	0.9894	0.2590
0.3937	18.28	16000	1.0197	0.2698

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1