ft_0117_korean

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: 0.8178
• Cer: 0.1942

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: 4
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
68.3088	0.03	100	125.2977	1.0
42.6071	0.07	200	75.7330	1.0
25.0826	0.1	300	35.9474	1.0
10.1653	0.14	400	7.3490	1.0
4.9322	0.17	500	5.3703	1.0
4.7973	0.21	600	5.2150	1.0
4.7712	0.24	700	5.1904	1.0
4.7076	0.28	800	5.0821	1.0
4.6923	0.31	900	4.9610	1.0
4.6677	0.34	1000	4.9497	1.0
4.7168	0.38	1100	5.0000	1.0
4.632	0.41	1200	4.9656	1.0
4.6183	0.45	1300	4.8375	1.0
4.5395	0.48	1400	4.6795	1.0
4.5385	0.52	1500	4.7110	1.0
4.5216	0.55	1600	4.6153	1.0
4.4442	0.59	1700	4.5397	1.0
4.3119	0.62	1800	4.2205	1.0
4.0832	0.66	1900	3.8678	0.9388
3.7389	0.69	2000	3.3075	0.6898
3.2707	0.72	2100	2.9589	0.5771
3.0099	0.76	2200	2.6719	0.5310
2.8581	0.79	2300	2.5290	0.5090
2.677	0.83	2400	2.3971	0.4737
2.5064	0.86	2500	2.2446	0.4643
2.4691	0.9	2600	2.1015	0.4373
2.2864	0.93	2700	2.0987	0.4290
2.2526	0.97	2800	2.0082	0.4223
2.156	1.0	2900	1.9078	0.4063
2.1556	1.03	3000	1.8029	0.3970
1.9783	1.07	3100	1.7724	0.3876
1.9678	1.1	3200	1.7115	0.3705
1.9242	1.14	3300	1.6834	0.3634
1.8216	1.17	3400	1.6559	0.3532
1.7855	1.21	3500	1.6106	0.3556
1.8123	1.24	3600	1.6309	0.3465
1.7609	1.28	3700	1.5353	0.3403
1.7131	1.31	3800	1.5067	0.3320
1.6954	1.35	3900	1.4273	0.3228
1.6219	1.38	4000	1.3992	0.3198
1.5606	1.41	4100	1.4247	0.3164
1.6549	1.45	4200	1.3775	0.3135
1.5869	1.48	4300	1.3162	0.3043
1.531	1.52	4400	1.2849	0.2980
1.4833	1.55	4500	1.3072	0.2989
1.4852	1.59	4600	1.2669	0.2984
1.4379	1.62	4700	1.2259	0.2897
1.4085	1.66	4800	1.2219	0.2828
1.4005	1.69	4900	1.1980	0.2791
1.3868	1.72	5000	1.2399	0.2874
1.3646	1.76	5100	1.2098	0.2829
1.3728	1.79	5200	1.2053	0.2779
1.2867	1.83	5300	1.1602	0.2737
1.3263	1.86	5400	1.1363	0.2650
1.2914	1.9	5500	1.1249	0.2611
1.2629	1.93	5600	1.0774	0.2559
1.2031	1.97	5700	1.1048	0.2570
1.2491	2.0	5800	1.0966	0.2655
1.159	2.04	5900	1.0593	0.2566
1.134	2.07	6000	1.0350	0.2475
1.1207	2.1	6100	1.0544	0.2455
1.116	2.14	6200	1.0340	0.2470
1.0947	2.17	6300	1.0177	0.2431
1.0844	2.21	6400	1.0166	0.2394
1.0679	2.24	6500	1.0666	0.2457
1.1139	2.28	6600	0.9607	0.2333
1.1074	2.31	6700	0.9982	0.2315
1.0263	2.35	6800	0.9937	0.2326
1.0264	2.38	6900	0.9521	0.2294
0.999	2.41	7000	0.9542	0.2306
1.0688	2.45	7100	0.9294	0.2251
1.0357	2.48	7200	0.9602	0.2231
1.0218	2.52	7300	0.9285	0.2264
0.9932	2.55	7400	0.9392	0.2224
1.0133	2.59	7500	0.9092	0.2225
1.0369	2.62	7600	0.9226	0.2185
0.9927	2.66	7700	0.9695	0.2236
1.0042	2.69	7800	0.9115	0.2200
0.9954	2.73	7900	0.8979	0.2151
0.9775	2.76	8000	0.9016	0.2161
0.9078	2.79	8100	0.9009	0.2177
0.9196	2.83	8200	0.9006	0.2149
0.9177	2.86	8300	0.8777	0.2125
0.8992	2.9	8400	0.8889	0.2097
0.911	2.93	8500	0.8693	0.2087
0.888	2.97	8600	0.8735	0.2134
0.9566	3.0	8700	0.8586	0.2078
0.8704	3.04	8800	0.8686	0.2079
0.8203	3.07	8900	0.8537	0.2064
0.8425	3.1	9000	0.8827	0.2065
0.8615	3.14	9100	0.8392	0.2030
0.8364	3.17	9200	0.8474	0.2041
0.8011	3.21	9300	0.8441	0.2031
0.8624	3.24	9400	0.8491	0.2046
0.8525	3.28	9500	0.8359	0.1995
0.8235	3.31	9600	0.8370	0.2008
0.8155	3.35	9700	0.8495	0.2015
0.7683	3.38	9800	0.8514	0.2014
0.8514	3.41	9900	0.8179	0.1981
0.8574	3.45	10000	0.8201	0.1992
0.819	3.48	10100	0.8437	0.1998
0.8078	3.52	10200	0.8292	0.1984
0.7888	3.55	10300	0.8253	0.1971
0.8105	3.59	10400	0.8180	0.1961
0.8006	3.62	10500	0.8111	0.1951
0.7583	3.66	10600	0.8265	0.1960
0.8425	3.69	10700	0.8248	0.1943
0.8564	3.73	10800	0.8250	0.1953
0.772	3.76	10900	0.8287	0.1970
0.8058	3.79	11000	0.8243	0.1961
0.7974	3.83	11100	0.8162	0.1944
0.7292	3.86	11200	0.8227	0.1950
0.7915	3.9	11300	0.8160	0.1941
0.7891	3.93	11400	0.8193	0.1945
0.7766	3.97	11500	0.8178	0.1942

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.1+cu121
• Datasets 2.13.0
• Tokenizers 0.15.0