ssc-hch-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: 2.4915
• Cer: 0.7560
• Wer: 1.0

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
5.8531	0.2506	100	3.0776	0.9917	1.0
3.1404	0.5013	200	2.9427	0.9917	1.0
3.5462	0.7519	300	3.0055	0.9917	1.0
3.3231	1.0025	400	2.9823	0.9917	1.0
3.2095	1.2531	500	2.9268	0.9917	1.0
3.1251	1.5038	600	2.9311	0.9917	1.0
3.1566	1.7544	700	2.9446	0.9917	1.0
3.1212	2.0050	800	2.9112	0.9917	1.0
3.0708	2.2556	900	2.9260	0.9829	1.0
3.0319	2.5063	1000	3.8026	0.9829	1.0
3.0086	2.7569	1100	3.3550	0.9829	1.0
2.9547	3.0075	1200	2.9463	0.9832	1.0
2.8971	3.2581	1300	3.1363	0.9661	1.0
2.8063	3.5088	1400	2.9367	0.9473	1.0
2.7296	3.7594	1500	3.0035	0.8580	1.0
2.5223	4.0100	1600	2.4125	0.7232	1.0
2.3708	4.2607	1700	2.4144	0.6792	1.0
2.3276	4.5113	1800	2.3535	0.6109	1.0
2.2372	4.7619	1900	2.1567	0.5972	1.0
2.136	5.0125	2000	2.2649	0.5880	0.9996
2.0764	5.2632	2100	2.0068	0.5701	1.0
2.0301	5.5138	2200	1.9794	0.5818	1.0
1.9486	5.7644	2300	2.0473	0.5687	0.9995
1.9485	6.0150	2400	1.9562	0.5288	1.0
1.8988	6.2657	2500	1.8791	0.5423	1.0
1.9147	6.5163	2600	1.8219	0.5453	1.0
1.8067	6.7669	2700	1.8768	0.5321	0.9998
1.7998	7.0175	2800	1.9343	0.5243	0.9996
1.7891	7.2682	2900	1.8222	0.5198	0.9998
1.7434	7.5188	3000	1.8008	0.5444	1.0
1.7674	7.7694	3100	1.9248	0.5005	0.9989
1.708	8.0201	3200	1.7520	0.5447	0.9995
1.7056	8.2707	3300	1.8185	0.5321	0.9995
1.6462	8.5213	3400	1.7989	0.5222	0.9998
1.6381	8.7719	3500	1.8162	0.5100	0.9986
1.6466	9.0226	3600	1.7556	0.5289	1.0
1.6018	9.2732	3700	1.6908	0.5010	0.9991
1.606	9.5238	3800	1.6967	0.4951	0.9977
1.6045	9.7744	3900	1.7846	0.5240	0.9982
1.5963	10.0251	4000	1.7657	0.5304	0.9995
1.5649	10.2757	4100	1.7817	0.5407	0.9995
1.5553	10.5263	4200	1.7213	0.5265	0.9993
1.5201	10.7769	4300	1.7401	0.5195	1.0
1.5254	11.0276	4400	1.8063	0.4984	0.9998
1.5068	11.2782	4500	1.8452	0.5403	1.0
1.5342	11.5288	4600	2.0170	0.5464	0.9998
1.7503	11.7794	4700	2.2264	0.5262	1.0
1.8808	12.0301	4800	1.9860	0.5120	1.0
1.9139	12.2807	4900	2.2726	0.5269	1.0
1.9509	12.5313	5000	1.9673	0.5404	1.0
2.0051	12.7820	5100	2.2250	0.5604	1.0
1.9579	13.0326	5200	2.1640	0.5959	1.0
1.8771	13.2832	5300	2.1336	0.5563	1.0
1.9823	13.5338	5400	2.2691	0.7077	1.0
2.1507	13.7845	5500	2.1997	0.6465	1.0
2.0604	14.0351	5600	2.3030	0.8980	1.0
2.1699	14.2857	5700	2.4337	0.8201	1.0
2.2419	14.5363	5800	2.4167	0.9188	1.0
2.2654	14.7870	5900	2.3791	0.9183	1.0
2.2122	15.0376	6000	2.3923	0.8131	1.0
2.1877	15.2882	6100	2.4024	0.7830	1.0
2.1916	15.5388	6200	2.4181	0.8381	1.0
2.1353	15.7895	6300	2.4418	0.8358	1.0
2.1799	16.0401	6400	2.4687	0.8174	1.0
2.2087	16.2907	6500	2.5634	0.8334	1.0
2.2825	16.5414	6600	2.5488	0.8351	1.0
2.3422	16.7920	6700	2.5349	0.8768	1.0
2.3002	17.0426	6800	2.5182	0.8676	1.0
2.3231	17.2932	6900	2.5707	0.8705	1.0
2.3212	17.5439	7000	2.5670	0.8576	1.0
2.294	17.7945	7100	2.5520	0.8529	1.0
2.3093	18.0451	7200	2.5077	0.8755	1.0
2.2777	18.2957	7300	2.5170	0.8419	1.0
2.2669	18.5464	7400	2.5190	0.8236	1.0
2.2838	18.7970	7500	2.4830	0.8454	1.0
2.2629	19.0476	7600	2.4891	0.8255	1.0
2.2429	19.2982	7700	2.4772	0.8228	1.0
2.2664	19.5489	7800	2.4824	0.8192	1.0
2.2216	19.7995	7900	2.4672	0.8239	1.0
2.2353	20.0501	8000	2.4799	0.7995	1.0
2.2275	20.3008	8100	2.4589	0.8008	1.0
2.2194	20.5514	8200	2.4732	0.7896	1.0
2.2165	20.8020	8300	2.4684	0.7895	1.0
2.1849	21.0526	8400	2.4801	0.7780	1.0
2.1829	21.3033	8500	2.4796	0.7700	1.0
2.1967	21.5539	8600	2.4524	0.7785	1.0
2.1703	21.8045	8700	2.4873	0.7599	1.0
2.195	22.0551	8800	2.4935	0.7567	1.0
2.2086	22.3058	8900	2.4447	0.7694	1.0
2.1866	22.5564	9000	2.4866	0.7501	1.0
2.1733	22.8070	9100	2.4694	0.7529	1.0
2.1637	23.0576	9200	2.4896	0.7430	1.0
2.1756	23.3083	9300	2.4785	0.7428	1.0
2.1634	23.5589	9400	2.5084	0.7336	1.0
2.1894	23.8095	9500	2.4891	0.7340	1.0
2.178	24.0602	9600	2.4982	0.7314	1.0
2.1725	24.3108	9700	2.4581	0.7392	1.0
2.1795	24.5614	9800	2.4721	0.7328	1.0
2.1679	24.8120	9900	2.4810	0.7300	1.0
2.2202	25.0627	10000	2.4784	0.7240	1.0
2.2071	25.3133	10100	2.5258	0.7131	1.0
2.2473	25.5639	10200	2.5072	0.7158	1.0
2.2474	25.8145	10300	2.5417	0.7117	1.0
2.2283	26.0652	10400	2.5465	0.7092	1.0
2.2543	26.3158	10500	2.5444	0.7102	1.0
2.2358	26.5664	10600	2.5406	0.7119	1.0
2.2889	26.8170	10700	2.4930	0.7231	1.0
2.2624	27.0677	10800	2.5067	0.7179	1.0
2.2619	27.3183	10900	2.4803	0.7338	1.0
2.2208	27.5689	11000	2.4662	0.7391	1.0
2.254	27.8195	11100	2.4554	0.7454	1.0
2.2567	28.0702	11200	2.4682	0.7418	1.0
2.2286	28.3208	11300	2.4771	0.7381	1.0
2.2648	28.5714	11400	2.4754	0.7405	1.0
2.2666	28.8221	11500	2.4745	0.7449	1.0
2.2456	29.0727	11600	2.4793	0.7509	1.0
2.2692	29.3233	11700	2.4868	0.7522	1.0
2.2602	29.5739	11800	2.4893	0.7541	1.0
2.2932	29.8246	11900	2.4915	0.7560	1.0

Framework versions

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