wav2vec2-xls-r-300m-CV-Fleurs-lg-50hrs-v6

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: 0.5412
• Wer: 0.3807
• Cer: 0.0816

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: 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
1.538	1.0	6320	0.7903	0.8894	0.2387
0.8457	2.0	12640	0.6261	0.8238	0.2064
0.6755	3.0	18960	0.4933	0.6752	0.1561
0.5729	4.0	25280	0.4407	0.5819	0.1323
0.5019	5.0	31600	0.4108	0.5538	0.1249
0.4479	6.0	37920	0.3908	0.5208	0.1158
0.4048	7.0	44240	0.3859	0.4932	0.1092
0.3658	8.0	50560	0.3573	0.4953	0.1105
0.3385	9.0	56880	0.3619	0.4860	0.1114
0.3135	10.0	63200	0.3545	0.4545	0.0995
0.2915	11.0	69520	0.3496	0.4523	0.0983
0.2721	12.0	75840	0.3473	0.4450	0.0982
0.2542	13.0	82160	0.3510	0.4327	0.0946
0.2379	14.0	88480	0.3679	0.4626	0.1020
0.2256	15.0	94800	0.3609	0.4410	0.0961
0.2114	16.0	101120	0.3616	0.4352	0.0947
0.2003	17.0	107440	0.3667	0.4333	0.0943
0.1897	18.0	113760	0.3805	0.4265	0.0926
0.1802	19.0	120080	0.3965	0.4197	0.0917
0.1704	20.0	126400	0.3803	0.4157	0.0903
0.1641	21.0	132720	0.3857	0.4298	0.0922
0.1559	22.0	139040	0.4016	0.4254	0.0925
0.1499	23.0	145360	0.4059	0.4133	0.0884
0.1434	24.0	151680	0.4082	0.4325	0.0915
0.138	25.0	158000	0.4483	0.4037	0.0888
0.1319	26.0	164320	0.4477	0.4197	0.0908
0.1294	27.0	170640	0.4377	0.4051	0.0880
0.1244	28.0	176960	0.4568	0.4252	0.0913
0.1206	29.0	183280	0.4550	0.4035	0.0872
0.1166	30.0	189600	0.4668	0.4074	0.0871
0.1118	31.0	195920	0.4386	0.3967	0.0867
0.1103	32.0	202240	0.4678	0.3915	0.0857
0.1077	33.0	208560	0.4852	0.3912	0.0852
0.1032	34.0	214880	0.4849	0.4003	0.0860
0.0993	35.0	221200	0.4955	0.3939	0.0841
0.0966	36.0	227520	0.4865	0.4001	0.0854
0.0945	37.0	233840	0.4700	0.3886	0.0843
0.0925	38.0	240160	0.4969	0.3801	0.0828
0.0898	39.0	246480	0.5060	0.3903	0.0837
0.0887	40.0	252800	0.5066	0.3910	0.0852
0.087	41.0	259120	0.4886	0.3827	0.0823
0.085	42.0	265440	0.5144	0.3899	0.0844
0.0825	43.0	271760	0.5123	0.3875	0.0838
0.0808	44.0	278080	0.5245	0.3820	0.0824
0.0791	45.0	284400	0.5231	0.3844	0.0822
0.0773	46.0	290720	0.5196	0.3867	0.0825
0.075	47.0	297040	0.5462	0.3839	0.0826
0.0737	48.0	303360	0.5412	0.3807	0.0816

Framework versions

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