xlsr-ewe

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2997
• Wer: 0.2635

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: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
169.4589	1.0	69	14.2354	1.0
60.2340	2.0	138	5.9731	1.0
31.7164	3.0	207	3.5702	1.0
25.3039	4.0	276	3.0823	1.0
24.5229	5.0	345	3.0302	1.0
24.5389	6.0	414	2.9540	1.0
17.9817	7.0	483	1.7917	0.9957
8.9278	8.0	552	0.8225	0.6369
7.1229	9.0	621	0.6046	0.5514
5.0350	10.0	690	0.5255	0.4525
5.6587	11.0	759	0.4622	0.4232
4.3718	12.0	828	0.4218	0.3913
4.2471	13.0	897	0.3807	0.3971
3.1870	14.0	966	0.3713	0.3589
3.6097	15.0	1035	0.3483	0.3485
3.1309	16.0	1104	0.3327	0.3340
2.9087	17.0	1173	0.3241	0.3249
2.7514	18.0	1242	0.3218	0.3244
2.8296	19.0	1311	0.3053	0.3123
2.2216	20.0	1380	0.3092	0.3032
2.5109	21.0	1449	0.2995	0.2971
2.7112	22.0	1518	0.3010	0.2964
2.0306	23.0	1587	0.3137	0.3006
2.2682	24.0	1656	0.3008	0.2847
2.2246	25.0	1725	0.3068	0.2889
2.0131	26.0	1794	0.2919	0.2999
2.3610	27.0	1863	0.2961	0.2884
1.7638	28.0	1932	0.2992	0.2799
2.1672	29.0	2001	0.2931	0.2786
1.6068	30.0	2070	0.2900	0.2706
1.6683	31.0	2139	0.3041	0.2693
1.6144	32.0	2208	0.2998	0.2752
1.5779	33.0	2277	0.2984	0.2676
1.4730	34.0	2346	0.2977	0.2643
1.2034	35.0	2415	0.3010	0.2615
1.3116	36.0	2484	0.3034	0.2685
1.4723	37.0	2553	0.3026	0.2661
1.1795	38.0	2622	0.3012	0.2654
1.2413	39.0	2691	0.2997	0.2622
1.2913	40.0	2760	0.2997	0.2635

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2