wav2vec2-xls-r-akan-100-hours

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: 0.7988
• Model Preparation Time: 0.0143
• Wer: 0.2968
• Cer: 0.0937

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: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer	Cer
11.1522	1.7331	500	2.7710	0.0143	1.0	1.0
2.0881	3.4662	1000	0.3882	0.0143	0.3401	0.1057
0.8886	5.1993	1500	0.3437	0.0143	0.2956	0.0916
0.7671	6.9324	2000	0.3246	0.0143	0.2898	0.0891
0.6983	8.6655	2500	0.3230	0.0143	0.2810	0.0872
0.6688	10.3986	3000	0.3235	0.0143	0.2800	0.0872
0.6241	12.1317	3500	0.3273	0.0143	0.2828	0.0879
0.5917	13.8648	4000	0.3328	0.0143	0.2836	0.0886
0.5503	15.5979	4500	0.3366	0.0143	0.2803	0.0882
0.5163	17.3310	5000	0.3568	0.0143	0.2825	0.0889
0.487	19.0641	5500	0.3597	0.0143	0.2876	0.0899
0.446	20.7972	6000	0.3719	0.0143	0.2831	0.0895
0.416	22.5303	6500	0.4071	0.0143	0.2964	0.0928
0.3844	24.2634	7000	0.4167	0.0143	0.2928	0.0924
0.3526	25.9965	7500	0.4353	0.0143	0.2999	0.0942
0.3173	27.7296	8000	0.4568	0.0143	0.3076	0.0968
0.2892	29.4627	8500	0.4936	0.0143	0.2990	0.0936
0.265	31.1958	9000	0.5298	0.0143	0.3044	0.0957
0.2452	32.9289	9500	0.5566	0.0143	0.2922	0.0930
0.2244	34.6620	10000	0.5921	0.0143	0.2973	0.0943
0.2064	36.3951	10500	0.6147	0.0143	0.3169	0.0980
0.1937	38.1282	11000	0.6672	0.0143	0.3118	0.0968
0.1733	39.8614	11500	0.6968	0.0143	0.2997	0.0938
0.1644	41.5945	12000	0.7098	0.0143	0.3010	0.0955
0.1527	43.3276	12500	0.7449	0.0143	0.2998	0.0947
0.1488	45.0607	13000	0.7555	0.0143	0.3054	0.0955
0.1341	46.7938	13500	0.7626	0.0143	0.3010	0.0951
0.1277	48.5269	14000	0.7988	0.0143	0.2968	0.0937

Framework versions

• Transformers 4.46.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.20.3