finetune

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: 0.2550
• Cer: 0.0688

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	Cer
10.1502	0.15	100	14.9196	0.9877
4.006	0.3	200	6.9761	1.0
3.2832	0.45	300	5.3599	1.0
3.2479	0.6	400	4.6352	0.9910
3.2182	0.74	500	4.6517	0.9397
3.1457	0.89	600	4.0926	0.9161
3.0171	1.04	700	3.5959	0.8029
2.7553	1.19	800	2.9406	0.7581
2.3743	1.34	900	1.8475	0.5338
2.0073	1.49	1000	1.2319	0.3782
1.8221	1.64	1100	0.9294	0.2806
1.6479	1.79	1200	1.0111	0.2248
1.5471	1.93	1300	0.8513	0.1568
1.4925	2.08	1400	0.5627	0.1451
1.429	2.23	1500	0.5325	0.1279
1.3717	2.38	1600	0.4783	0.1149
1.3732	2.53	1700	0.4600	0.1111
1.3256	2.68	1800	0.4764	0.1027
1.3044	2.83	1900	0.4076	0.1012
1.2801	2.98	2000	0.4318	0.1007
1.2196	3.12	2100	0.3867	0.1005
1.232	3.27	2200	0.4147	0.0948
1.2354	3.42	2300	0.3613	0.0939
1.2679	3.57	2400	0.3724	0.0941
1.1958	3.72	2500	0.3626	0.0908
1.2021	3.87	2600	0.3800	0.0886
1.2087	4.02	2700	0.3640	0.0879
1.1263	4.17	2800	0.3576	0.0879
1.1494	4.32	2900	0.3402	0.0838
1.0579	4.46	3000	0.3286	0.0847
1.0966	4.61	3100	0.3234	0.0849
1.1303	4.76	3200	0.3244	0.0841
1.121	4.91	3300	0.3040	0.0821
1.0772	5.06	3400	0.3252	0.0843
1.092	5.21	3500	0.3142	0.0818
1.1267	5.36	3600	0.3844	0.0848
1.0902	5.51	3700	0.3079	0.0807
1.0584	5.65	3800	0.3337	0.0825
1.0457	5.8	3900	0.3302	0.0830
1.0282	5.95	4000	0.3056	0.0813
1.0741	6.1	4100	0.2867	0.0793
0.9822	6.25	4200	0.3002	0.0805
1.0194	6.4	4300	0.2873	0.0781
1.0271	6.55	4400	0.2861	0.0785
1.04	6.7	4500	0.2881	0.0787
1.0276	6.85	4600	0.2763	0.0781
1.0476	6.99	4700	0.2911	0.0791
0.989	7.14	4800	0.2947	0.0807
1.0077	7.29	4900	0.2905	0.0779
1.0095	7.44	5000	0.2883	0.0786
0.9498	7.59	5100	0.2823	0.0778
0.9677	7.74	5200	0.2870	0.0783
0.9795	7.89	5300	0.2813	0.0769
0.9706	8.04	5400	0.2771	0.0753
0.9582	8.18	5500	0.2798	0.0752
0.975	8.33	5600	0.2935	0.0778
0.9624	8.48	5700	0.2827	0.0777
0.9646	8.63	5800	0.2747	0.0741
1.0015	8.78	5900	0.2752	0.0745
0.9449	8.93	6000	0.2740	0.0771
0.9205	9.08	6100	0.2793	0.0773
0.9554	9.23	6200	0.2839	0.0761
0.948	9.38	6300	0.2715	0.0755
0.9308	9.52	6400	0.2772	0.0768
0.9227	9.67	6500	0.2751	0.0759
0.9908	9.82	6600	0.2655	0.0767
0.974	9.97	6700	0.2752	0.0745
0.8731	10.12	6800	0.2739	0.0747
0.9545	10.27	6900	0.2747	0.0744
0.894	10.42	7000	0.2755	0.0752
0.9346	10.57	7100	0.2772	0.0744
0.9275	10.71	7200	0.2714	0.0737
0.899	10.86	7300	0.2747	0.0743
0.947	11.01	7400	0.2714	0.0748
0.8735	11.16	7500	0.2691	0.0731
0.9134	11.31	7600	0.2737	0.0734
0.9061	11.46	7700	0.2812	0.0750
0.9179	11.61	7800	0.2731	0.0742
0.8899	11.76	7900	0.2716	0.0739
0.8736	11.9	8000	0.2706	0.0735
0.9004	12.05	8100	0.2755	0.0747
0.8915	12.2	8200	0.2798	0.0740
0.8572	12.35	8300	0.2739	0.0743
0.8512	12.5	8400	0.2759	0.0758
0.8617	12.65	8500	0.2715	0.0745
0.9042	12.8	8600	0.2668	0.0726
0.8908	12.95	8700	0.2728	0.0738
0.9157	13.1	8800	0.2672	0.0715
0.8568	13.24	8900	0.2738	0.0742
0.8354	13.39	9000	0.2706	0.0726
0.8462	13.54	9100	0.2681	0.0730
0.854	13.69	9200	0.2703	0.0737
0.8584	13.84	9300	0.2663	0.0735
0.8378	13.99	9400	0.2666	0.0739
0.8656	14.14	9500	0.2694	0.0744
0.8072	14.29	9600	0.2742	0.0733
0.8369	14.43	9700	0.2667	0.0735
0.8587	14.58	9800	0.2660	0.0732
0.8227	14.73	9900	0.2625	0.0739
0.8624	14.88	10000	0.2661	0.0731
0.8515	15.03	10100	0.2643	0.0721
0.8688	15.18	10200	0.2664	0.0723
0.8469	15.33	10300	0.2612	0.0713
0.874	15.48	10400	0.2656	0.0720
0.84	15.62	10500	0.2623	0.0722
0.8408	15.77	10600	0.2625	0.0717
0.8419	15.92	10700	0.2619	0.0720
0.8177	16.07	10800	0.2620	0.0716
0.8168	16.22	10900	0.2687	0.0727
0.8347	16.37	11000	0.2635	0.0713
0.8161	16.52	11100	0.2598	0.0709
0.7783	16.67	11200	0.2705	0.0729
0.8253	16.82	11300	0.2618	0.0722
0.8604	16.96	11400	0.2688	0.0734
0.7786	17.11	11500	0.2654	0.0727
0.8296	17.26	11600	0.2669	0.0724
0.805	17.41	11700	0.2667	0.0723
0.7961	17.56	11800	0.2636	0.0725
0.8497	17.71	11900	0.2626	0.0718
0.8123	17.86	12000	0.2636	0.0720
0.7842	18.01	12100	0.2644	0.0720
0.8391	18.15	12200	0.2629	0.0720
0.8324	18.3	12300	0.2656	0.0725
0.8114	18.45	12400	0.2642	0.0714
0.8014	18.6	12500	0.2611	0.0716
0.771	18.75	12600	0.2601	0.0721
0.7998	18.9	12700	0.2606	0.0715
0.7253	19.05	12800	0.2617	0.0718
0.8057	19.2	12900	0.2607	0.0718
0.817	19.35	13000	0.2614	0.0720

Framework versions

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