wav2vec2-xls-r-300m-Fleurs_AMMI_AFRIVOICE_LRSC-ln-10hrs-v2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4351
• Wer: 0.2577
• Cer: 0.0833

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: 1e-05
• train_batch_size: 4
• eval_batch_size: 2
• seed: 42
• optimizer: Use 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: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
7.7647	1.0	768	3.1933	1.0	1.0
3.0093	2.0	1536	2.8646	1.0	1.0
2.8666	3.0	2304	2.8188	1.0	1.0
2.7248	4.0	3072	2.2736	1.0	0.7102
1.8518	5.0	3840	1.1205	0.9523	0.2684
1.2788	6.0	4608	0.8644	0.6379	0.1739
1.0685	7.0	5376	0.7393	0.5324	0.1498
0.9306	8.0	6144	0.6640	0.4750	0.1369
0.8376	9.0	6912	0.6044	0.4290	0.1282
0.7758	10.0	7680	0.5699	0.4055	0.1215
0.7303	11.0	8448	0.5588	0.3849	0.1159
0.6748	12.0	9216	0.5249	0.3716	0.1115
0.6509	13.0	9984	0.5161	0.3550	0.1074
0.6141	14.0	10752	0.5091	0.3599	0.1076
0.5903	15.0	11520	0.4849	0.3486	0.1050
0.5676	16.0	12288	0.4614	0.3460	0.1053
0.5421	17.0	13056	0.4494	0.3476	0.1080
0.5363	18.0	13824	0.4598	0.3189	0.0980
0.5139	19.0	14592	0.4553	0.3157	0.0976
0.5002	20.0	15360	0.4492	0.3118	0.0960
0.4847	21.0	16128	0.4415	0.3145	0.0972
0.467	22.0	16896	0.4287	0.2988	0.0923
0.4558	23.0	17664	0.4367	0.2970	0.0927
0.4442	24.0	18432	0.4265	0.3019	0.0951
0.4274	25.0	19200	0.4247	0.2910	0.0912
0.424	26.0	19968	0.4205	0.2900	0.0915
0.403	27.0	20736	0.4131	0.2909	0.0915
0.4017	28.0	21504	0.4068	0.3002	0.0985
0.393	29.0	22272	0.4118	0.2895	0.0910
0.3852	30.0	23040	0.4275	0.2857	0.0897
0.3783	31.0	23808	0.4145	0.2852	0.0911
0.3776	32.0	24576	0.4200	0.2781	0.0882
0.3667	33.0	25344	0.4174	0.2778	0.0876
0.3582	34.0	26112	0.4129	0.2827	0.0902
0.3517	35.0	26880	0.4231	0.2780	0.0870
0.3479	36.0	27648	0.4058	0.2806	0.0910
0.3394	37.0	28416	0.4375	0.2757	0.0868
0.3336	38.0	29184	0.4320	0.2723	0.0861
0.3267	39.0	29952	0.4165	0.2750	0.0871
0.3179	40.0	30720	0.4198	0.2729	0.0865
0.3157	41.0	31488	0.4152	0.2721	0.0861
0.3097	42.0	32256	0.4007	0.2753	0.0886
0.306	43.0	33024	0.4471	0.2705	0.0853
0.3061	44.0	33792	0.4154	0.2649	0.0849
0.2931	45.0	34560	0.4506	0.2611	0.0832
0.2994	46.0	35328	0.4287	0.2669	0.0851
0.2972	47.0	36096	0.4188	0.2670	0.0859
0.2836	48.0	36864	0.4350	0.2554	0.0828
0.283	49.0	37632	0.4167	0.2620	0.0845
0.2765	50.0	38400	0.4314	0.2634	0.0849
0.278	51.0	39168	0.4097	0.2702	0.0892
0.2744	52.0	39936	0.4255	0.2591	0.0835
0.2662	53.0	40704	0.4293	0.2595	0.0836
0.2663	54.0	41472	0.4423	0.2583	0.0826
0.264	55.0	42240	0.4313	0.2587	0.0835
0.2549	56.0	43008	0.4441	0.2614	0.0842
0.2586	57.0	43776	0.4259	0.2611	0.0857
0.2577	58.0	44544	0.4258	0.2570	0.0837
0.2486	59.0	45312	0.4332	0.2606	0.0845
0.2473	60.0	46080	0.4402	0.2589	0.0834
0.2464	61.0	46848	0.4263	0.2567	0.0826
0.2442	62.0	47616	0.4527	0.2563	0.0819
0.2444	63.0	48384	0.4351	0.2577	0.0833

Framework versions

• Transformers 4.46.1
• Pytorch 2.1.0+cu118
• Datasets 3.1.0
• Tokenizers 0.20.1