es_ar

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4940
• Wer: 0.3263
• Cer: 0.2414

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 15.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.6158	0.4	500	3.7004	0.9886	0.9625
1.0731	0.8	1000	1.0090	0.7501	0.4066
0.7529	1.2	1500	0.7609	0.6195	0.3460
0.6441	1.6	2000	0.6322	0.5524	0.3171
0.607	2.0	2500	0.5795	0.5202	0.3065
0.4744	2.4	3000	0.5848	0.5096	0.3056
0.4604	2.8	3500	0.5341	0.4666	0.2907
0.3763	3.2	4000	0.5060	0.4416	0.2812
0.3952	3.6	4500	0.5214	0.4566	0.2850
0.3962	4.0	5000	0.4890	0.4324	0.2784
0.3137	4.4	5500	0.4833	0.4165	0.2713
0.316	4.8	6000	0.5005	0.4182	0.2738
0.2721	5.2	6500	0.4961	0.4171	0.2732
0.2561	5.6	7000	0.4742	0.3997	0.2645
0.2854	6.0	7500	0.4600	0.3991	0.2662
0.2599	6.4	8000	0.4541	0.4022	0.2659
0.2249	6.8	8500	0.4586	0.3911	0.2615
0.1931	7.2	9000	0.4721	0.3871	0.2614
0.195	7.6	9500	0.4636	0.3898	0.2608
0.1991	8.0	10000	0.4259	0.3716	0.2555
0.1657	8.4	10500	0.4548	0.3714	0.2573
0.1802	8.8	11000	0.4540	0.3582	0.2526
0.1359	9.2	11500	0.4685	0.3652	0.2552
0.1419	9.6	12000	0.4524	0.3561	0.2512
0.1531	10.0	12500	0.4443	0.3578	0.2514
0.1313	10.4	13000	0.4536	0.3536	0.2495
0.1269	10.8	13500	0.4563	0.3517	0.2480
0.102	11.2	14000	0.4606	0.3424	0.2476
0.103	11.6	14500	0.4611	0.3489	0.2477
0.1088	12.0	15000	0.4505	0.3362	0.2447
0.0917	12.4	15500	0.4741	0.3404	0.2458
0.0847	12.8	16000	0.4714	0.3340	0.2440
0.0768	13.2	16500	0.4943	0.3286	0.2427
0.0789	13.6	17000	0.4813	0.3308	0.2429
0.0797	14.0	17500	0.4861	0.3288	0.2423
0.0698	14.4	18000	0.5003	0.3271	0.2416
0.0686	14.8	18500	0.4940	0.3263	0.2414

Framework versions

• Transformers 4.36.2
• Pytorch 1.13.0+cu116
• Datasets 2.15.0
• Tokenizers 0.15.2