torgo_xlsr_finetune-M05-2

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: 1.5128
• Wer: 1.1148

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
22.2047	0.86	500	3.3446	0.9930
3.3679	1.72	1000	2.9591	0.9930
2.8813	2.58	1500	2.7978	0.9930
2.7207	3.44	2000	2.5604	0.9930
2.3274	4.3	2500	2.0135	1.4468
1.5821	5.16	3000	1.6148	1.5686
1.1549	6.02	3500	1.3447	1.5014
0.8908	6.88	4000	1.3315	1.4524
0.7204	7.75	4500	1.3250	1.3894
0.6209	8.61	5000	1.2566	1.3697
0.5507	9.47	5500	1.2300	1.3221
0.4622	10.33	6000	1.3826	1.3165
0.4503	11.19	6500	1.2769	1.2717
0.4026	12.05	7000	1.3531	1.2955
0.3617	12.91	7500	1.2806	1.2521
0.3239	13.77	8000	1.5507	1.2437
0.3051	14.63	8500	1.6217	1.2563
0.2983	15.49	9000	1.5210	1.2185
0.2766	16.35	9500	1.4787	1.2143
0.2642	17.21	10000	1.6284	1.2311
0.2358	18.07	10500	1.3203	1.1891
0.2303	18.93	11000	1.5233	1.2185
0.2166	19.79	11500	1.5111	1.2129
0.2162	20.65	12000	1.5128	1.1919
0.1981	21.51	12500	1.4668	1.1877
0.1736	22.38	13000	1.5041	1.1485
0.1725	23.24	13500	1.5306	1.1639
0.1632	24.1	14000	1.3756	1.1373
0.1597	24.96	14500	1.5404	1.1345
0.1571	25.82	15000	1.4863	1.1359
0.1569	26.68	15500	1.4775	1.1401
0.1431	27.54	16000	1.5410	1.1218
0.1373	28.4	16500	1.5212	1.1246
0.1461	29.26	17000	1.5128	1.1148

Framework versions

• Transformers 4.26.1
• Pytorch 1.13.1+cu116
• Datasets 1.18.3
• Tokenizers 0.13.2