working

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: 0.2518
• Wer: 0.5902
• Cer: 0.0676

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: 3e-05
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2000
• training_steps: 13000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
10.4868	0.15	100	13.8019	1.0	0.9317
4.7063	0.3	200	9.0935	1.0	0.9692
3.624	0.45	300	6.7956	1.0	0.9692
3.5139	0.6	400	4.9444	1.0	0.9692
3.4538	0.74	500	4.4408	1.0	0.8260
3.393	0.89	600	4.1983	1.0	0.8183
3.3836	1.04	700	3.9939	1.0	0.7414
3.3494	1.19	800	3.8468	1.0	0.7642
3.2885	1.34	900	3.4016	1.0	0.7987
3.2538	1.49	1000	3.1741	1.0	0.8071
3.2633	1.64	1100	3.1409	1.0008	0.7878
3.1524	1.79	1200	2.9351	1.0	0.7935
3.0518	1.93	1300	2.7497	1.0	0.7559
2.807	2.08	1400	2.0357	0.9992	0.6716
2.3203	2.23	1500	1.2459	0.9927	0.4256
1.9018	2.38	1600	0.8000	0.9513	0.2472
1.6926	2.53	1700	0.6204	0.9066	0.1800
1.5501	2.68	1800	0.5426	0.8619	0.1377
1.4654	2.83	1900	0.4881	0.8050	0.1199
1.4087	2.98	2000	0.4747	0.7961	0.1153
1.3228	3.12	2100	0.4334	0.7953	0.1138
1.3292	3.27	2200	0.4214	0.7604	0.1072
1.3123	3.42	2300	0.3850	0.7400	0.0998
1.3417	3.57	2400	0.4069	0.7595	0.1039
1.2557	3.72	2500	0.3867	0.7246	0.0970
1.2588	3.87	2600	0.3678	0.7311	0.0952
1.2602	4.02	2700	0.3682	0.7262	0.0937
1.1774	4.17	2800	0.3530	0.7043	0.0908
1.2075	4.32	2900	0.3343	0.7076	0.0885
1.1085	4.46	3000	0.3542	0.7124	0.0898
1.1451	4.61	3100	0.3325	0.6872	0.0870
1.1721	4.76	3200	0.3318	0.6848	0.0858
1.1549	4.91	3300	0.3145	0.6816	0.0862
1.1276	5.06	3400	0.3133	0.6613	0.0822
1.1193	5.21	3500	0.3117	0.6783	0.0832
1.158	5.36	3600	0.3111	0.6856	0.0841
1.1235	5.51	3700	0.3130	0.6856	0.0848
1.1075	5.65	3800	0.3144	0.6621	0.0831
1.0834	5.8	3900	0.3228	0.6694	0.0873
1.0751	5.95	4000	0.3116	0.6734	0.0823
1.1114	6.1	4100	0.3022	0.6718	0.0812
1.0025	6.25	4200	0.3149	0.6807	0.0826
1.0488	6.4	4300	0.2957	0.6377	0.0799
1.0602	6.55	4400	0.3002	0.6604	0.0800
1.0694	6.7	4500	0.3003	0.6580	0.0799
1.0583	6.85	4600	0.2850	0.6182	0.0777
1.0737	6.99	4700	0.2876	0.6288	0.0785
1.0123	7.14	4800	0.3064	0.6791	0.0806
1.0305	7.29	4900	0.2925	0.6271	0.0796
1.0269	7.44	5000	0.2978	0.6271	0.0796
0.9634	7.59	5100	0.2943	0.6353	0.0791
0.986	7.74	5200	0.2947	0.6556	0.0795
1.0039	7.89	5300	0.2891	0.6580	0.0793
0.9928	8.04	5400	0.2901	0.6426	0.0786
0.9799	8.18	5500	0.2827	0.6385	0.0775
0.9952	8.33	5600	0.3063	0.6677	0.0791
0.9874	8.48	5700	0.2922	0.6401	0.0780
0.9716	8.63	5800	0.2806	0.6060	0.0748
1.0133	8.78	5900	0.2824	0.6328	0.0759
0.9665	8.93	6000	0.2849	0.6442	0.0782
0.9516	9.08	6100	0.2795	0.6158	0.0766
0.9864	9.23	6200	0.2998	0.6336	0.0788
0.9652	9.38	6300	0.2812	0.6174	0.0760
0.9518	9.52	6400	0.2824	0.6182	0.0764
0.9447	9.67	6500	0.2872	0.6434	0.0777
1.0079	9.82	6600	0.2668	0.6158	0.0753
0.9996	9.97	6700	0.2721	0.6084	0.0749
0.9042	10.12	6800	0.2755	0.6109	0.0746
0.965	10.27	6900	0.2747	0.6093	0.0751
0.9172	10.42	7000	0.2799	0.6247	0.0763
0.9441	10.57	7100	0.2833	0.6231	0.0771
0.9417	10.71	7200	0.2736	0.6117	0.0754
0.9091	10.86	7300	0.2783	0.6182	0.0760
0.96	11.01	7400	0.2674	0.6149	0.0759
0.8866	11.16	7500	0.2703	0.6198	0.0749
0.9357	11.31	7600	0.2796	0.6117	0.0769
0.9144	11.46	7700	0.2743	0.6084	0.0753
0.9456	11.61	7800	0.2675	0.6206	0.0750
0.9022	11.76	7900	0.2680	0.6093	0.0736
0.9047	11.9	8000	0.2753	0.6158	0.0766
0.9151	12.05	8100	0.2772	0.6328	0.0761
0.9123	12.2	8200	0.2879	0.6247	0.0758
0.8731	12.35	8300	0.2764	0.6093	0.0743
0.8745	12.5	8400	0.2796	0.6214	0.0751
0.8812	12.65	8500	0.2684	0.6060	0.0739
0.9228	12.8	8600	0.2678	0.6149	0.0744
0.9029	12.95	8700	0.2718	0.5963	0.0738
0.934	13.1	8800	0.2700	0.6174	0.0737
0.8622	13.24	8900	0.2789	0.6231	0.0757
0.848	13.39	9000	0.2725	0.6060	0.0735
0.8727	13.54	9100	0.2602	0.6133	0.0737
0.8705	13.69	9200	0.2661	0.6247	0.0748
0.8724	13.84	9300	0.2658	0.6190	0.0741
0.8609	13.99	9400	0.2677	0.6320	0.0753
0.8836	14.14	9500	0.2692	0.6353	0.0753
0.8419	14.29	9600	0.2715	0.6158	0.0734
0.8583	14.43	9700	0.2596	0.6060	0.0726
0.882	14.58	9800	0.2636	0.6198	0.0749
0.8396	14.73	9900	0.2627	0.6125	0.0739
0.8832	14.88	10000	0.2676	0.6271	0.0739
0.8647	15.03	10100	0.2669	0.6190	0.0742
0.8674	15.18	10200	0.2677	0.6158	0.0741
0.8621	15.33	10300	0.2605	0.6028	0.0711
0.8946	15.48	10400	0.2630	0.6036	0.0722
0.8544	15.62	10500	0.2644	0.6133	0.0737
0.8684	15.77	10600	0.2624	0.5963	0.0731
0.861	15.92	10700	0.2626	0.5987	0.0735
0.8322	16.07	10800	0.2635	0.6003	0.0736
0.8277	16.22	10900	0.2729	0.6214	0.0742
0.8463	16.37	11000	0.2663	0.6068	0.0739
0.8506	16.52	11100	0.2629	0.6003	0.0728
0.8005	16.67	11200	0.2657	0.6206	0.0748
0.8336	16.82	11300	0.2625	0.6003	0.0721
0.8676	16.96	11400	0.2691	0.6174	0.0733
0.8028	17.11	11500	0.2635	0.6003	0.0720
0.85	17.26	11600	0.2602	0.6011	0.0726
0.8366	17.41	11700	0.2658	0.6068	0.0729
0.8098	17.56	11800	0.2621	0.6003	0.0720
0.8666	17.71	11900	0.2600	0.5963	0.0734
0.8271	17.86	12000	0.2618	0.6076	0.0734
0.8018	18.01	12100	0.2669	0.6117	0.0737
0.8471	18.15	12200	0.2630	0.6028	0.0727
0.8573	18.3	12300	0.2666	0.6117	0.0735
0.8331	18.45	12400	0.2639	0.6060	0.0727
0.8161	18.6	12500	0.2606	0.6044	0.0731
0.8021	18.75	12600	0.2591	0.6019	0.0726
0.8236	18.9	12700	0.2611	0.6036	0.0722
0.7332	19.05	12800	0.2610	0.6060	0.0732
0.8267	19.2	12900	0.2599	0.6028	0.0733
0.8296	19.35	13000	0.2608	0.6036	0.0731

Framework versions

• Transformers 4.17.0
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0