wav2vec2-adafactor

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on the None dataset. It achieves the following results on the evaluation set:

• Loss: 521.5777
• Wer: 0.5680
• Cer: 0.2057

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
6431.0881	0.17	200	8827.3496	1.0	0.9509
2376.9597	0.34	400	2278.7505	1.0	1.0
1966.3091	0.52	600	1954.2694	1.0	1.0
1921.7344	0.69	800	1877.0073	1.0	1.0
1904.6434	0.86	1000	1867.8109	1.0	1.0
2007.2327	1.03	1200	1853.8639	1.0	1.0
1820.0666	1.21	1400	1840.9535	1.0	1.0
1788.3647	1.38	1600	1708.9249	1.0	0.9990
1445.4336	1.55	1800	1239.5427	0.9994	0.5676
1263.3165	1.72	2000	1028.6985	0.9856	0.4438
1166.8841	1.89	2200	919.8116	0.9513	0.3761
1083.2934	2.07	2400	878.6225	0.9199	0.3472
1018.7722	2.24	2600	838.0692	0.9111	0.3313
1024.2178	2.41	2800	803.7117	0.8689	0.3172
934.4987	2.58	3000	801.0446	0.8736	0.3370
896.6591	2.76	3200	769.0162	0.8474	0.3170
853.1276	2.93	3400	754.6758	0.8170	0.2946
877.0165	3.1	3600	747.9600	0.8277	0.3014
895.9198	3.27	3800	725.9247	0.8085	0.3023
865.6773	3.44	4000	764.1653	0.8372	0.3611
829.4269	3.62	4200	714.3458	0.7792	0.2840
837.7436	3.79	4400	691.1579	0.7660	0.2776
840.2796	3.96	4600	662.4272	0.7488	0.2704
780.3309	4.13	4800	644.3304	0.7395	0.2723
808.9362	4.3	5000	644.2058	0.7362	0.2688
833.0216	4.48	5200	660.3318	0.7217	0.2586
688.053	4.65	5400	640.7992	0.7176	0.2575
734.9365	4.82	5600	644.1985	0.7103	0.2536
686.6702	4.99	5800	621.6605	0.7017	0.2513
783.8035	5.17	6000	609.5834	0.6936	0.2485
767.5277	5.34	6200	607.1106	0.6841	0.2438
774.2312	5.51	6400	621.5826	0.6810	0.2433
711.3146	5.68	6600	589.4904	0.6740	0.2419
715.3734	5.85	6800	588.7719	0.6695	0.2401
741.9265	6.03	7000	615.9477	0.6577	0.2336
725.2223	6.2	7200	590.8420	0.6560	0.2332
753.4423	6.37	7400	557.4046	0.6555	0.2404
672.7011	6.54	7600	585.4437	0.6401	0.2296
660.0451	6.72	7800	587.1536	0.6348	0.2268
685.1729	6.89	8000	566.8586	0.6402	0.2307
707.4157	7.06	8200	564.1030	0.6312	0.2285
648.6515	7.23	8400	570.9037	0.6287	0.2253
671.3892	7.4	8600	569.9574	0.6261	0.2254
654.9685	7.58	8800	548.4863	0.6181	0.2244
668.527	7.75	9000	561.7392	0.6167	0.2208
618.0621	7.92	9200	568.9872	0.6172	0.2209
679.1587	8.09	9400	548.2069	0.6114	0.2208
664.1059	8.27	9600	554.3419	0.6122	0.2207
642.3993	8.44	9800	555.8182	0.6080	0.2192
634.971	8.61	10000	552.2766	0.6029	0.2158
657.4219	8.78	10200	545.1649	0.6014	0.2153
625.1877	8.95	10400	551.3602	0.6009	0.2162
593.6537	9.13	10600	556.0078	0.5980	0.2157
631.1515	9.3	10800	552.9581	0.5987	0.2145
628.705	9.47	11000	542.7684	0.5991	0.2155
669.4295	9.64	11200	549.7736	0.6008	0.2166
615.2577	9.81	11400	545.1942	0.5966	0.2134
666.2597	9.99	11600	540.7983	0.5909	0.2121
568.6661	10.16	11800	540.6088	0.5939	0.2125
599.7543	10.33	12000	542.0428	0.5905	0.2129
660.4966	10.5	12200	531.7808	0.5925	0.2143
605.1542	10.68	12400	545.8644	0.5944	0.2138
629.2883	10.85	12600	543.6411	0.5839	0.2096
602.7661	11.02	12800	538.6114	0.5780	0.2089
600.5382	11.19	13000	533.9185	0.5891	0.2135
626.9337	11.36	13200	531.1143	0.5788	0.2088
566.7313	11.54	13400	529.5084	0.5819	0.2095
689.2272	11.71	13600	530.3742	0.5844	0.2107
613.5476	11.88	13800	527.6745	0.5800	0.2093
639.2343	12.05	14000	536.8987	0.5745	0.2078
613.0707	12.23	14200	534.7601	0.5761	0.2085
617.9929	12.4	14400	527.2368	0.5760	0.2082
603.2728	12.57	14600	520.8328	0.5730	0.2075
589.526	12.74	14800	527.3028	0.5704	0.2063
613.8767	12.91	15000	520.7419	0.5693	0.2059
526.2904	13.09	15200	526.8046	0.5683	0.2055
597.296	13.26	15400	523.0710	0.5724	0.2069
582.0505	13.43	15600	522.9764	0.5688	0.2057
637.9532	13.6	15800	523.2065	0.5701	0.2063
609.4863	13.78	16000	524.9271	0.5699	0.2059
630.7555	13.95	16200	522.3800	0.5702	0.2063
578.2891	14.12	16400	525.6607	0.5704	0.2062
642.7179	14.29	16600	522.1490	0.5706	0.2066
565.622	14.46	16800	524.6316	0.5666	0.2052
605.6358	14.64	17000	523.2621	0.5684	0.2056
608.121	14.81	17200	522.6130	0.5687	0.2058
556.9755	14.98	17400	521.5777	0.5680	0.2057

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.0+cu118
• Datasets 2.16.0
• Tokenizers 0.15.2