Modelpublic

faixalsarah/wav2vec2-xlsr-dhivehi-finetuned-v2

sourceHugging Faceapache-2.0updated 8mo agoView on Hugging Face

0likes52downloads

Model Card

wav2vec2-xlsr-dhivehi-finetuned-v2

This model is a fine-tuned version of sammy786/wav2vec2-xlsr-dhivehi on the None dataset. It achieves the following results on the evaluation set:

—Loss: inf
—Wer: 0.4562

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: 4.5e-05
—trainbatchsize: 8
—evalbatchsize: 16
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 16
—optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: cosinewithrestarts
—lrschedulerwarmup_steps: 500
—num_epochs: 30
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4775	0.8045	500	0.3821	0.5334
0.4568	1.6082	1000	0.3337	0.4920
0.4331	2.4119	1500	0.3119	0.4996
0.3714	3.2156	2000	0.3188	0.4897
0.4034	4.0193	2500	0.2970	0.5033
0.3334	4.8238	3000	0.2862	0.4780
0.3404	5.6275	3500	0.2991	0.4774
0.3087	6.4312	4000	0.2818	0.4708
0.302	7.2349	4500	0.3048	0.4769
0.3179	8.0386	5000	0.2899	0.4765
0.2628	8.8431	5500	0.2986	0.4701
0.2395	9.6468	6000	0.3026	0.4758
0.2451	10.4505	6500	0.2882	0.4717
0.2298	11.2542	7000	0.2923	0.4678
0.2477	12.0579	7500	0.3163	0.4654
0.2178	12.8624	8000	0.3080	0.4636
0.2215	13.6661	8500	0.2907	0.4741
0.2229	14.4698	9000	0.3145	0.4688
0.2087	15.2735	9500	0.2927	0.4713
0.2426	16.0772	10000	0.3167	0.4745
0.2852	16.8817	10500	0.3132	0.4652
0.2797	17.6854	11000	0.3147	0.5195
0.3735	18.4891	11500	0.3181	0.5022

Framework versions

—Transformers 4.57.3
—Pytorch 2.9.0+cu126
—Datasets 4.0.0
—Tokenizers 0.22.2