ssc-koo-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: 2.2968
• Cer: 0.6783
• Wer: 0.9996

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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
5.3204	0.1813	100	3.0099	0.9947	1.0
3.0302	0.3626	200	2.9541	0.9947	1.0
2.9391	0.5440	300	2.9475	0.9891	1.0
2.8252	0.7253	400	2.8337	0.9449	1.0
2.6387	0.9066	500	2.8135	0.8428	1.0
2.4564	1.0870	600	2.4902	0.7996	1.0
2.2494	1.2684	700	2.3222	0.7157	1.0
2.0578	1.4497	800	2.2174	0.6497	1.0
1.96	1.6310	900	2.1959	0.6285	1.0
1.9028	1.8123	1000	2.0725	0.6030	1.0
1.861	1.9937	1100	2.1194	0.5821	1.0
1.7362	2.1741	1200	2.0431	0.5334	1.0034
1.7346	2.3554	1300	1.9258	0.5612	0.9989
1.6872	2.5367	1400	2.0139	0.5290	0.9994
1.6921	2.7180	1500	1.8111	0.5355	0.9975
1.6316	2.8994	1600	1.8895	0.5355	0.9979
1.6122	3.0798	1700	1.7664	0.5184	0.9977
1.5547	3.2611	1800	1.7894	0.5413	0.9987
1.522	3.4424	1900	1.8288	0.5181	0.9981
1.5581	3.6238	2000	1.7491	0.5186	0.9977
1.5435	3.8051	2100	1.6929	0.4941	0.9958
1.4729	3.9864	2200	1.7429	0.5246	0.9989
1.4626	4.1668	2300	1.8013	0.4950	1.0098
1.4534	4.3481	2400	1.7123	0.4895	0.9938
1.4282	4.5295	2500	1.6176	0.5090	0.9977
1.4085	4.7108	2600	1.7041	0.4888	0.9945
1.4031	4.8921	2700	1.6435	0.4764	0.9915
1.3729	5.0725	2800	1.8482	0.4806	1.0034
1.3016	5.2539	2900	1.6509	0.4884	0.9932
1.3634	5.4352	3000	1.6329	0.4900	0.9920
1.3245	5.6165	3100	1.6840	0.4913	1.0036
1.3025	5.7978	3200	1.7043	0.5168	1.0015
1.4095	5.9791	3300	1.6779	0.5539	0.9966
1.2452	6.1596	3400	1.7308	0.4870	0.9934
1.2521	6.3409	3500	1.5969	0.4836	0.9903
1.2549	6.5222	3600	1.6128	0.4952	0.9907
1.2759	6.7035	3700	1.5583	0.4781	0.9924
1.2093	6.8849	3800	1.6189	0.4503	1.0220
1.2453	7.0653	3900	1.5713	0.4635	1.0824
1.147	7.2466	4000	1.6146	0.5098	1.0038
1.1986	7.4279	4100	1.6173	0.4708	1.0062
1.1709	7.6092	4200	1.7183	0.4605	1.0021
1.1371	7.7906	4300	1.5477	0.5110	0.9932
1.2075	7.9719	4400	1.5412	0.4962	0.9913
1.1176	8.1523	4500	1.5634	0.4670	0.9936
1.107	8.3336	4600	1.5297	0.5283	0.9883
1.0861	8.5150	4700	1.5289	0.4419	0.9775
1.1011	8.6963	4800	1.5835	0.4770	0.9890
1.0864	8.8776	4900	1.5043	0.4636	0.9841
1.0748	9.0580	5000	1.6871	0.5933	0.9962
1.0034	9.2393	5100	1.5045	0.4699	0.9879
1.0074	9.4207	5200	1.5527	0.4470	1.0078
1.02	9.6020	5300	1.5285	0.4432	0.9835
1.0147	9.7833	5400	1.5470	0.4755	0.9911
1.0409	9.9646	5500	1.5250	0.4645	0.9852
0.9529	10.1451	5600	1.5696	0.4458	1.0038
0.9414	10.3264	5700	1.5319	0.4810	0.9858
0.941	10.5077	5800	1.5790	0.4538	1.0044
0.9145	10.6890	5900	1.5246	0.4837	0.9873
0.9824	10.8704	6000	1.4837	0.4579	0.9890
0.9397	11.0508	6100	1.4778	0.4493	0.9797
0.8304	11.2321	6200	1.5285	0.4179	1.0068
0.8759	11.4134	6300	1.5563	0.4266	0.9871
0.8738	11.5947	6400	1.5636	0.4260	0.9938
0.8908	11.7761	6500	1.5436	0.4385	0.9792
0.8985	11.9574	6600	1.5019	0.4239	0.9818
0.8332	12.1378	6700	1.5872	0.4421	1.0106
0.7945	12.3191	6800	1.5605	0.4375	0.9915
0.815	12.5005	6900	1.5049	0.4306	0.9737
0.8124	12.6818	7000	1.5192	0.4348	0.9778
0.8422	12.8631	7100	1.5167	0.4165	0.9748
0.7935	13.0435	7200	1.5640	0.4264	0.9826
0.7438	13.2248	7300	1.6165	0.4271	0.9881
0.7894	13.4062	7400	1.5718	0.4120	0.9663
0.7721	13.5875	7500	1.6259	0.4228	0.9818
0.7704	13.7688	7600	1.5615	0.4129	0.9716
0.77	13.9501	7700	1.5601	0.4167	0.9756
0.693	14.1306	7800	1.8324	0.4425	1.1045
0.7443	14.3119	7900	1.6445	0.4302	1.0028
0.731	14.4932	8000	1.7441	0.4224	0.9839
0.7405	14.6745	8100	1.6907	0.4266	1.0367
0.7987	14.8558	8200	1.6870	0.4337	1.0163
0.7793	15.0363	8300	1.6622	0.4282	0.9716
0.7963	15.2176	8400	1.6690	0.4499	1.0193
0.8667	15.3989	8500	1.7338	0.4621	1.1327
0.9406	15.5802	8600	1.9090	0.4674	0.9909
1.026	15.7616	8700	1.9508	0.4935	0.9879
1.082	15.9429	8800	1.8690	0.4628	0.9896
1.0817	16.1233	8900	1.9416	0.5220	0.9911
1.129	16.3046	9000	1.9048	0.4781	1.0136
1.1041	16.4859	9100	1.9063	0.4870	1.0098
1.0897	16.6673	9200	1.6682	0.4857	0.9833
0.8812	16.8486	9300	1.5728	0.4479	0.9947
0.8284	17.0290	9400	1.7669	0.4586	1.1358
0.8431	17.2103	9500	1.6971	0.4425	1.0203
0.8755	17.3917	9600	1.7615	0.4822	0.9879
1.062	17.5730	9700	1.7330	0.5176	0.9869
1.2399	17.7543	9800	1.9085	0.5983	0.9943
1.397	17.9356	9900	1.9057	0.6441	1.0
1.471	18.1160	10000	1.9789	0.6564	1.0
1.5444	18.2974	10100	2.0396	0.7628	1.0
1.6848	18.4787	10200	2.2329	0.8280	1.0
1.9732	18.6600	10300	2.4373	0.9427	1.0
2.0928	18.8413	10400	2.6932	0.9935	1.0
2.3268	19.0218	10500	2.7731	0.9921	1.0
2.3669	19.2031	10600	2.7349	0.9937	1.0
2.3452	19.3844	10700	2.6973	0.9898	1.0
2.3012	19.5657	10800	2.6725	0.9906	1.0
2.2682	19.7471	10900	2.6361	0.9805	1.0
2.2306	19.9284	11000	2.6518	0.9856	1.0
2.1568	20.1088	11100	2.6378	0.9823	1.0
2.0884	20.2901	11200	2.6507	0.9789	1.0
2.043	20.4714	11300	2.5891	0.9679	1.0
1.9813	20.6528	11400	2.5639	0.9523	1.0
1.9409	20.8341	11500	2.5460	0.9400	1.0
1.8849	21.0145	11600	2.6583	0.9198	1.0
1.8339	21.1958	11700	2.5035	0.8808	1.0
1.88	21.3772	11800	2.4978	0.8844	1.0
1.8623	21.5585	11900	2.4886	0.8847	1.0
1.8547	21.7398	12000	2.4982	0.8833	1.0
1.8228	21.9211	12100	2.4950	0.8875	1.0
1.7964	22.1015	12200	2.5017	0.8840	1.0
1.8215	22.2829	12300	2.4629	0.8650	1.0
1.7828	22.4642	12400	2.4534	0.8535	1.0
1.7645	22.6455	12500	2.4305	0.8543	1.0
1.7831	22.8268	12600	2.4408	0.8496	1.0
1.7373	23.0073	12700	2.4332	0.8323	1.0
1.7207	23.1886	12800	2.4330	0.8385	1.0
1.745	23.3699	12900	2.4564	0.8564	1.0
1.711	23.5512	13000	2.4077	0.8317	1.0
1.6738	23.7325	13100	2.4319	0.8411	1.0
1.6978	23.9139	13200	2.3909	0.8197	1.0
1.6858	24.0943	13300	2.4286	0.8360	1.0
1.6762	24.2756	13400	2.3942	0.8206	1.0
1.6632	24.4569	13500	2.4042	0.8250	1.0
1.6843	24.6383	13600	2.3702	0.8026	1.0
1.6441	24.8196	13700	2.3766	0.8090	1.0
1.645	25.0	13800	2.3549	0.7979	1.0
1.6432	25.1813	13900	2.3644	0.8056	1.0
1.6348	25.3626	14000	2.3551	0.7923	1.0
1.6151	25.5440	14100	2.3470	0.7952	1.0
1.6106	25.7253	14200	2.3524	0.7957	1.0
1.6091	25.9066	14300	2.3157	0.7692	1.0
1.6046	26.0870	14400	2.3130	0.7698	1.0
1.6114	26.2684	14500	2.3034	0.7634	1.0
1.6161	26.4497	14600	2.2931	0.7479	1.0
1.5859	26.6310	14700	2.2782	0.7385	1.0
1.5979	26.8123	14800	2.2741	0.7380	1.0
1.6055	26.9937	14900	2.2754	0.7348	1.0
1.5975	27.1741	15000	2.2700	0.7280	1.0
1.5867	27.3554	15100	2.2661	0.7300	1.0
1.5912	27.5367	15200	2.2641	0.7236	1.0
1.5794	27.7180	15300	2.2615	0.7243	1.0
1.6202	27.8994	15400	2.2637	0.7218	1.0
1.6007	28.0798	15500	2.2634	0.7072	1.0
1.5792	28.2611	15600	2.2635	0.7044	1.0
1.6341	28.4424	15700	2.2641	0.7010	1.0
1.5974	28.6238	15800	2.2650	0.6961	1.0
1.5844	28.8051	15900	2.2757	0.6931	0.9998
1.6409	28.9864	16000	2.2818	0.6873	0.9998
1.6391	29.1668	16100	2.2872	0.6828	0.9996
1.6373	29.3481	16200	2.2886	0.6816	0.9996
1.6202	29.5295	16300	2.2917	0.6805	0.9996
1.6358	29.7108	16400	2.2953	0.6781	0.9996
1.6427	29.8921	16500	2.2968	0.6783	0.9996

Framework versions

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