wav2vec2-xls-r-300m-Fleurs_AMMI_AFRIVOICE_LRSC-ln-100hrs-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: 0.3710
• Wer: 0.1939
• Cer: 0.0632

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
3.0206	1.0	5760	0.6687	0.4122	0.1239
0.8506	2.0	11520	0.4709	0.3106	0.0945
0.7012	3.0	17280	0.3870	0.3028	0.0920
0.6373	4.0	23040	0.3576	0.2734	0.0826
0.598	5.0	28800	0.3538	0.2662	0.0814
0.5728	6.0	34560	0.3209	0.2538	0.0781
0.5507	7.0	40320	0.3151	0.2467	0.0751
0.5327	8.0	46080	0.3041	0.2341	0.0727
0.517	9.0	51840	0.3150	0.2384	0.0736
0.5056	10.0	57600	0.3031	0.2215	0.0697
0.4968	11.0	63360	0.2960	0.2184	0.0684
0.4829	12.0	69120	0.3008	0.2166	0.0679
0.4736	13.0	74880	0.2885	0.2153	0.0667
0.4628	14.0	80640	0.3040	0.2144	0.0669
0.4556	15.0	86400	0.2832	0.2049	0.0638
0.4435	16.0	92160	0.3059	0.2063	0.0646
0.4358	17.0	97920	0.2956	0.2079	0.0658
0.4244	18.0	103680	0.2960	0.2019	0.0628
0.4137	19.0	109440	0.2882	0.2066	0.0655
0.4054	20.0	115200	0.2945	0.2033	0.0658
0.3997	21.0	120960	0.2998	0.2061	0.0657
0.3897	22.0	126720	0.2866	0.2025	0.0669
0.3779	23.0	132480	0.3029	0.2013	0.0685
0.3686	24.0	138240	0.3168	0.1965	0.0632
0.3619	25.0	144000	0.3174	0.1972	0.0648
0.3549	26.0	149760	0.3099	0.1994	0.0656
0.3483	27.0	155520	0.3107	0.2015	0.0694
0.3401	28.0	161280	0.3254	0.1934	0.0626
0.3319	29.0	167040	0.3291	0.1990	0.0663
0.3268	30.0	172800	0.3275	0.1941	0.0656
0.3209	31.0	178560	0.3213	0.2037	0.0700
0.3157	32.0	184320	0.3189	0.1994	0.0675
0.3103	33.0	190080	0.3433	0.1952	0.0618
0.3063	34.0	195840	0.3200	0.1934	0.0629
0.3009	35.0	201600	0.3270	0.1940	0.0655
0.2966	36.0	207360	0.3395	0.1926	0.0628
0.2921	37.0	213120	0.3619	0.1893	0.0612
0.2893	38.0	218880	0.3535	0.1923	0.0642
0.285	39.0	224640	0.3570	0.1942	0.0649
0.2796	40.0	230400	0.3634	0.1890	0.0612
0.2759	41.0	236160	0.3369	0.1995	0.0669
0.2743	42.0	241920	0.3605	0.1945	0.0655
0.27	43.0	247680	0.3838	0.1902	0.0627
0.2676	44.0	253440	0.3750	0.1895	0.0628
0.2643	45.0	259200	0.3672	0.1917	0.0637
0.2603	46.0	264960	0.3717	0.1922	0.0641
0.2555	47.0	270720	0.3710	0.1939	0.0632

Framework versions

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