wav2vec2-xls-r-300m-CV_Fleurs_AMMI_ALFFA-sw-20hrs-v1

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: 1.1489
• Wer: 0.2863
• Cer: 0.0861

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: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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
5.8088	0.9993	752	1.5829	0.9976	0.5075
2.8319	2.0	1505	1.1351	0.8587	0.2839
2.2211	2.9993	2257	0.9053	0.7716	0.2313
1.8858	4.0	3010	0.8276	0.6929	0.2025
1.6621	4.9993	3762	0.7863	0.6353	0.1842
1.4776	6.0	4515	0.7335	0.5902	0.1687
1.33	6.9993	5267	0.7035	0.5585	0.1592
1.1967	8.0	6020	0.6806	0.5353	0.1533
1.0763	8.9993	6772	0.6649	0.5126	0.1482
0.9834	10.0	7525	0.6615	0.4852	0.1400
0.9072	10.9993	8277	0.6760	0.4776	0.1374
0.8453	12.0	9030	0.6869	0.4568	0.1311
0.7749	12.9993	9782	0.6667	0.4582	0.1339
0.7246	14.0	10535	0.6778	0.4384	0.1290
0.6844	14.9993	11287	0.6678	0.4239	0.1225
0.6404	16.0	12040	0.6619	0.4218	0.1229
0.5983	16.9993	12792	0.7178	0.4176	0.1223
0.5783	18.0	13545	0.7103	0.4068	0.1172
0.5414	18.9993	14297	0.7358	0.4018	0.1166
0.5223	20.0	15050	0.7178	0.4028	0.1148
0.5012	20.9993	15802	0.7341	0.3924	0.1153
0.4729	22.0	16555	0.7488	0.3965	0.1145
0.4672	22.9993	17307	0.7782	0.3830	0.1115
0.4397	24.0	18060	0.7940	0.3715	0.1078
0.4339	24.9993	18812	0.7789	0.3830	0.1108
0.4164	26.0	19565	0.7889	0.3904	0.1130
0.4085	26.9993	20317	0.7793	0.3794	0.1101
0.3883	28.0	21070	0.7890	0.3680	0.1073
0.383	28.9993	21822	0.8251	0.3691	0.1076
0.3676	30.0	22575	0.8064	0.3675	0.1075
0.3531	30.9993	23327	0.8620	0.3669	0.1084
0.3391	32.0	24080	0.8385	0.3597	0.1060
0.3358	32.9993	24832	0.8355	0.3609	0.1075
0.3284	34.0	25585	0.8700	0.3704	0.1089
0.3189	34.9993	26337	0.8719	0.3628	0.1080
0.3086	36.0	27090	0.8334	0.3629	0.1055
0.3084	36.9993	27842	0.8646	0.3496	0.1036
0.2911	38.0	28595	0.8664	0.3560	0.1044
0.2887	38.9993	29347	0.9033	0.3467	0.1019
0.279	40.0	30100	0.8944	0.3500	0.1031
0.277	40.9993	30852	0.8604	0.3503	0.1029
0.2643	42.0	31605	0.8859	0.3459	0.1023
0.2623	42.9993	32357	0.9263	0.3394	0.1003
0.2558	44.0	33110	0.9256	0.3387	0.1000
0.2527	44.9993	33862	0.9429	0.3370	0.1000
0.2439	46.0	34615	0.9764	0.3383	0.1001
0.2377	46.9993	35367	0.9389	0.3330	0.0983
0.2434	48.0	36120	0.9903	0.3325	0.0986
0.2271	48.9993	36872	0.9591	0.3305	0.0986
0.2219	50.0	37625	0.9128	0.3310	0.0981
0.2176	50.9993	38377	0.9322	0.3310	0.0978
0.2174	52.0	39130	0.9558	0.3282	0.0967
0.2205	52.9993	39882	0.9752	0.3281	0.0972
0.2011	54.0	40635	1.0161	0.3296	0.0974
0.2027	54.9993	41387	1.0044	0.3231	0.0957
0.2007	56.0	42140	0.9645	0.3203	0.0955
0.1958	56.9993	42892	1.0168	0.3202	0.0953
0.1908	58.0	43645	1.0060	0.3241	0.0960
0.1874	58.9993	44397	0.9889	0.3209	0.0953
0.1832	60.0	45150	0.9944	0.3218	0.0951
0.1835	60.9993	45902	1.0524	0.3195	0.0948
0.1821	62.0	46655	1.0135	0.3189	0.0946
0.1746	62.9993	47407	0.9728	0.3158	0.0931
0.1714	64.0	48160	1.0200	0.3128	0.0931
0.1677	64.9993	48912	0.9743	0.3123	0.0928
0.1654	66.0	49665	1.0243	0.3137	0.0931
0.1649	66.9993	50417	1.0631	0.3149	0.0939
0.1574	68.0	51170	1.0807	0.3139	0.0935
0.155	68.9993	51922	1.0730	0.3100	0.0919
0.1576	70.0	52675	1.0481	0.3111	0.0927
0.1529	70.9993	53427	1.0492	0.3092	0.0920
0.1518	72.0	54180	1.0321	0.3062	0.0916
0.1455	72.9993	54932	1.0690	0.3032	0.0914
0.1428	74.0	55685	1.0508	0.3003	0.0903
0.1399	74.9993	56437	1.0759	0.3050	0.0910
0.1384	76.0	57190	1.0915	0.2993	0.0906
0.1348	76.9993	57942	1.1089	0.3015	0.0903
0.1346	78.0	58695	1.1002	0.3024	0.0902
0.1302	78.9993	59447	1.0900	0.3004	0.0897
0.1294	80.0	60200	1.0783	0.2995	0.0891
0.1266	80.9993	60952	1.0696	0.2966	0.0884
0.121	82.0	61705	1.1130	0.2957	0.0881
0.1212	82.9993	62457	1.0869	0.2960	0.0888
0.1203	84.0	63210	1.1091	0.2921	0.0882
0.1173	84.9993	63962	1.1155	0.2943	0.0885
0.1148	86.0	64715	1.1247	0.2954	0.0884
0.1122	86.9993	65467	1.1156	0.2942	0.0879
0.1215	88.0	66220	1.1048	0.2948	0.0880
0.1143	88.9993	66972	1.1065	0.2946	0.0879
0.1114	90.0	67725	1.1122	0.2919	0.0875
0.1067	90.9993	68477	1.1245	0.2938	0.0878
0.1056	92.0	69230	1.1245	0.2918	0.0872
0.1084	92.9993	69982	1.1358	0.2899	0.0870
0.104	94.0	70735	1.1432	0.2882	0.0867
0.1019	94.9993	71487	1.1435	0.2877	0.0862
0.0993	96.0	72240	1.1500	0.2870	0.0860
0.1012	96.9993	72992	1.1342	0.2855	0.0859
0.1023	98.0	73745	1.1397	0.2855	0.0859
0.0962	98.9993	74497	1.1491	0.2855	0.0859
0.0954	99.9336	75200	1.1489	0.2863	0.0861

Framework versions

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