wav2vec2-large-xlsr-53_train_data_full

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4168
• Wer: 0.3383

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.0459	0.73	500	3.2037	0.9995
0.7938	1.45	1000	0.7432	0.6373
0.503	2.18	1500	0.5517	0.5115
0.4475	2.91	2000	0.4916	0.4624
0.3575	3.63	2500	0.4612	0.4362
0.3206	4.36	3000	0.4546	0.4198
0.3155	5.09	3500	0.4073	0.3929
0.2827	5.81	4000	0.4172	0.3808
0.2575	6.54	4500	0.4183	0.3741
0.2399	7.27	5000	0.4181	0.3680
0.2455	7.99	5500	0.3981	0.3604
0.2512	8.72	6000	0.4203	0.3612
0.221	9.45	6500	0.4073	0.3560
0.19	10.17	7000	0.4206	0.3547
0.207	10.9	7500	0.3992	0.3517
0.187	11.63	8000	0.4078	0.3517
0.2029	12.35	8500	0.4143	0.3469
0.171	13.08	9000	0.4007	0.3430
0.1658	13.81	9500	0.3862	0.3422
0.2021	14.53	10000	0.4132	0.3454
0.165	15.26	10500	0.3997	0.3407
0.1562	15.99	11000	0.4069	0.3416
0.1613	16.71	11500	0.4040	0.3393
0.1713	17.44	12000	0.4094	0.3411
0.1541	18.17	12500	0.4043	0.3367
0.144	18.89	13000	0.4086	0.3374
0.1483	19.62	13500	0.4168	0.3383

Framework versions

• Transformers 4.18.0
• Pytorch 1.11.0+cu102
• Datasets 2.1.0
• Tokenizers 0.12.1