wav2vecresultsfinale

This model is a fine-tuned version of anish-shilpakar/wav2vec2-nepali on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3374
• Wer: 0.2672

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 16
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.274	0.2584	100	2.9770	1.0
2.6375	0.5168	200	2.5025	0.9899
2.0193	0.7752	300	1.7123	0.9546
1.2825	1.0336	400	1.1216	0.7681
1.1884	1.2920	500	0.8725	0.6704
0.935	1.5504	600	0.6984	0.6106
0.8219	1.8088	700	0.6194	0.5110
0.947	2.0672	800	0.5441	0.4978
0.7252	2.3256	900	0.5025	0.4827
0.6017	2.5840	1000	0.4948	0.4436
0.5668	2.8424	1100	0.4635	0.4279
0.5788	3.1008	1200	0.4433	0.4159
0.6228	3.3592	1300	0.4221	0.3882
0.5826	3.6176	1400	0.4101	0.3819
0.5042	3.8760	1500	0.4021	0.3724
0.5962	4.1344	1600	0.3826	0.3699
0.4774	4.3928	1700	0.3832	0.3472
0.3943	4.6512	1800	0.3757	0.3554
0.5177	4.9096	1900	0.3702	0.3636
0.4252	5.1680	2000	0.3784	0.3239
0.4051	5.4264	2100	0.3765	0.3233
0.4709	5.6848	2200	0.3668	0.3352
0.5013	5.9432	2300	0.3479	0.3258
0.4049	6.2016	2400	0.3518	0.3151
0.396	6.4599	2500	0.3450	0.3157
0.3179	6.7183	2600	0.3462	0.3012
0.4026	6.9767	2700	0.3413	0.2987
0.3065	7.2351	2800	0.3466	0.3012
0.3578	7.4935	2900	0.3485	0.2936
0.3801	7.7519	3000	0.3481	0.2917
0.2889	8.0103	3100	0.3349	0.3006
0.3422	8.2687	3200	0.3473	0.2905
0.2781	8.5271	3300	0.3429	0.3018
0.2766	8.7855	3400	0.3447	0.3031
0.2966	9.0439	3500	0.3416	0.3043
0.3177	9.3023	3600	0.3366	0.2892
0.3245	9.5607	3700	0.3421	0.2924
0.3776	9.8191	3800	0.3477	0.2962
0.2868	10.0775	3900	0.3309	0.2899
0.3264	10.3359	4000	0.3317	0.2905
0.3352	10.5943	4100	0.3363	0.2911
0.2949	10.8527	4200	0.3304	0.2823
0.2962	11.1111	4300	0.3335	0.2829
0.2903	11.3695	4400	0.3364	0.2817
0.2602	11.6279	4500	0.3380	0.2823
0.3312	11.8863	4600	0.3309	0.2760
0.2659	12.1447	4700	0.3316	0.2798
0.3107	12.4031	4800	0.3421	0.2842
0.3231	12.6615	4900	0.3328	0.2766
0.2178	12.9199	5000	0.3308	0.2710
0.2772	13.1783	5100	0.3382	0.2691
0.2842	13.4367	5200	0.3348	0.2773
0.2871	13.6951	5300	0.3346	0.2678
0.3136	13.9535	5400	0.3345	0.2703
0.2265	14.2119	5500	0.3362	0.2703
0.2409	14.4703	5600	0.3358	0.2716
0.3904	14.7287	5700	0.3365	0.2697
0.2499	14.9871	5800	0.3338	0.2691
0.2285	15.2455	5900	0.3347	0.2653
0.2518	15.5039	6000	0.3372	0.2653
0.2218	15.7623	6100	0.3374	0.2672

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.2