ssc-bxk-model

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.4388
• Cer: 0.2469
• Wer: 0.8253

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
4.6605	0.2882	100	3.0343	0.9955	1.0
2.9398	0.5764	200	2.9498	0.9919	0.9998
2.6066	0.8646	300	2.6204	0.7305	1.0
1.927	1.1527	400	1.6190	0.4561	0.9996
1.5776	1.4409	500	1.5378	0.4264	0.9989
1.438	1.7291	600	1.2293	0.3542	0.9624
1.3485	2.0173	700	1.1700	0.3398	0.9502
1.2258	2.3055	800	1.1301	0.3231	0.9289
1.2223	2.5937	900	1.1717	0.4102	0.9601
1.2276	2.8818	1000	1.0739	0.3068	0.9080
1.1002	3.1700	1100	1.0304	0.3090	0.9174
1.1105	3.4582	1200	0.9805	0.3034	0.9022
1.0848	3.7464	1300	1.0215	0.3204	0.9146
1.0122	4.0346	1400	1.0029	0.2903	0.8837
0.9599	4.3228	1500	0.9633	0.2795	0.8630
0.9635	4.6110	1600	1.0093	0.3015	0.8924
0.9947	4.8991	1700	1.0203	0.2732	0.8813
0.9357	5.1873	1800	1.0053	0.2866	0.8841
0.8969	5.4755	1900	0.9909	0.2936	0.8892
0.8773	5.7637	2000	0.9700	0.3065	0.8843
0.8913	6.0519	2100	0.9634	0.2909	0.8829
0.819	6.3401	2200	0.9919	0.2963	0.8823
0.8203	6.6282	2300	0.9623	0.2711	0.8655
0.8374	6.9164	2400	0.9565	0.2723	0.8690
0.7793	7.2046	2500	0.9639	0.2778	0.8531
0.7874	7.4928	2600	0.9875	0.2762	0.8756
0.7736	7.7810	2700	0.9054	0.2717	0.8355
0.7649	8.0692	2800	0.9827	0.2746	0.8425
0.7166	8.3573	2900	0.9269	0.2601	0.8293
0.726	8.6455	3000	0.9668	0.2670	0.8520
0.6946	8.9337	3100	0.9494	0.2612	0.8423
0.6848	9.2219	3200	1.0094	0.2536	0.8348
0.6637	9.5101	3300	0.9842	0.2664	0.8412
0.6711	9.7983	3400	0.9495	0.2477	0.8239
0.6417	10.0865	3500	0.9724	0.2724	0.8511
0.6206	10.3746	3600	0.9884	0.2635	0.8373
0.6172	10.6628	3700	1.0208	0.2652	0.8504
0.6383	10.9510	3800	0.9368	0.2594	0.8343
0.5773	11.2392	3900	1.0053	0.2598	0.8499
0.561	11.5274	4000	0.9953	0.2579	0.8301
0.5901	11.8156	4100	0.9727	0.2612	0.8338
0.572	12.1037	4200	1.0612	0.2633	0.8458
0.531	12.3919	4300	0.9773	0.2610	0.8441
0.5407	12.6801	4400	0.9359	0.2546	0.8079
0.5481	12.9683	4500	0.9912	0.2622	0.8423
0.5121	13.2565	4600	1.0189	0.2435	0.8021
0.4812	13.5447	4700	1.0280	0.2644	0.8283
0.4957	13.8329	4800	1.0750	0.2621	0.8418
0.4883	14.1210	4900	1.0284	0.2574	0.8182
0.4452	14.4092	5000	1.0755	0.2676	0.8336
0.4784	14.6974	5100	1.0643	0.2469	0.8194
0.4816	14.9856	5200	1.0728	0.2671	0.8410
0.4237	15.2738	5300	1.0982	0.2646	0.8460
0.4245	15.5620	5400	1.0886	0.2393	0.8038
0.4351	15.8501	5500	1.0750	0.2522	0.8471
0.4229	16.1383	5600	1.1751	0.2555	0.8288
0.4111	16.4265	5700	1.1463	0.2584	0.8357
0.3947	16.7147	5800	1.1446	0.2586	0.8290
0.4146	17.0029	5900	1.1675	0.2507	0.8256
0.3463	17.2911	6000	1.1274	0.2540	0.8386
0.3682	17.5793	6100	1.1349	0.2604	0.8278
0.361	17.8674	6200	1.1877	0.2584	0.8449
0.3672	18.1556	6300	1.1213	0.2570	0.8324
0.3533	18.4438	6400	1.0914	0.2516	0.8079
0.3433	18.7320	6500	1.1553	0.2489	0.8155
0.3431	19.0202	6600	1.1193	0.2487	0.8201
0.3138	19.3084	6700	1.1092	0.2462	0.8116
0.329	19.5965	6800	1.1028	0.2475	0.8129
0.3017	19.8847	6900	1.2623	0.2516	0.8200
0.2905	20.1729	7000	1.1891	0.2584	0.8336
0.2828	20.4611	7100	1.1295	0.2531	0.8081
0.2975	20.7493	7200	1.1500	0.2608	0.8299
0.2898	21.0375	7300	1.1343	0.2500	0.8263
0.2694	21.3256	7400	1.2274	0.2489	0.8228
0.2694	21.6138	7500	1.2690	0.2515	0.8272
0.2672	21.9020	7600	1.2604	0.2574	0.8246
0.2505	22.1902	7700	1.2891	0.2496	0.8192
0.246	22.4784	7800	1.2759	0.2547	0.8198
0.2651	22.7666	7900	1.2823	0.2520	0.8293
0.2407	23.0548	8000	1.3137	0.2506	0.8329
0.2312	23.3429	8100	1.3152	0.2587	0.8347
0.2227	23.6311	8200	1.3156	0.2495	0.8159
0.2152	23.9193	8300	1.3123	0.2484	0.8327
0.2137	24.2075	8400	1.3881	0.2477	0.8210
0.2096	24.4957	8500	1.3389	0.2525	0.8230
0.2167	24.7839	8600	1.3801	0.2499	0.8249
0.215	25.0720	8700	1.4032	0.2524	0.8301
0.1984	25.3602	8800	1.4061	0.2495	0.8292
0.2002	25.6484	8900	1.3790	0.2483	0.8311
0.1921	25.9366	9000	1.3252	0.2464	0.8239
0.185	26.2248	9100	1.4067	0.2482	0.8301
0.1718	26.5130	9200	1.3845	0.2466	0.8173
0.1881	26.8012	9300	1.3831	0.2474	0.8173
0.1795	27.0893	9400	1.4146	0.2489	0.8150
0.1655	27.3775	9500	1.4118	0.2482	0.8153
0.1722	27.6657	9600	1.3973	0.2465	0.8208
0.1736	27.9539	9700	1.3843	0.2476	0.8205
0.1656	28.2421	9800	1.4340	0.2470	0.8256
0.1644	28.5303	9900	1.4053	0.2461	0.8240
0.1613	28.8184	10000	1.4068	0.2472	0.8217
0.1639	29.1066	10100	1.4115	0.2443	0.8198
0.1537	29.3948	10200	1.4257	0.2462	0.8219
0.1549	29.6830	10300	1.4378	0.2464	0.8258
0.1448	29.9712	10400	1.4388	0.2469	0.8253

Framework versions

• Transformers 4.57.2
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.22.0