ssc-cgg-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: 1.2753
• Cer: 0.2625
• Wer: 0.8266

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
4.774	0.1864	100	2.9668	0.9919	1.0
3.2708	0.3728	200	2.9400	0.9919	1.0
2.8413	0.5592	300	2.8321	0.9919	1.0
2.674	0.7456	400	2.6069	0.9055	1.0
2.4214	0.9320	500	2.3585	0.7714	1.0
2.0978	1.1174	600	1.8840	0.6053	1.0
1.9201	1.3038	700	1.7315	0.5596	0.9999
1.8358	1.4902	800	1.7703	0.5090	0.9960
1.7211	1.6766	900	1.5671	0.4943	0.9927
1.6456	1.8630	1000	1.5215	0.4965	0.9921
1.6238	2.0485	1100	1.5559	0.5125	0.9971
1.4901	2.2349	1200	1.4645	0.4661	0.9902
1.5001	2.4212	1300	1.4230	0.4627	0.9884
1.4129	2.6076	1400	1.4149	0.4542	0.9870
1.4087	2.7940	1500	1.3809	0.4403	0.9839
1.3885	2.9804	1600	1.4125	0.4351	0.9868
1.3267	3.1659	1700	1.3841	0.4168	1.0238
1.2669	3.3523	1800	1.4178	0.4564	0.9826
1.2492	3.5387	1900	1.3101	0.4087	0.9711
1.242	3.7251	2000	1.2938	0.4311	0.9801
1.2144	3.9115	2100	1.3929	0.4172	0.9994
1.2334	4.0969	2200	1.2291	0.3985	0.9631
1.1318	4.2833	2300	1.2567	0.4069	0.9707
1.1588	4.4697	2400	1.2027	0.3812	0.9539
1.1203	4.6561	2500	1.3124	0.4197	0.9724
1.117	4.8425	2600	1.1737	0.3909	0.9588
1.0582	5.0280	2700	1.1263	0.3584	0.9352
0.9921	5.2144	2800	1.1103	0.3723	0.9463
1.0199	5.4007	2900	1.0575	0.3422	0.9376
1.0181	5.5871	3000	1.1219	0.3554	0.9393
1.0148	5.7735	3100	1.0687	0.3575	0.9324
1.0273	5.9599	3200	1.1047	0.3482	0.9379
0.9599	6.1454	3300	1.0959	0.3559	1.0169
0.9171	6.3318	3400	1.0382	0.3395	0.9403
0.8826	6.5182	3500	1.0086	0.3233	0.9215
0.9008	6.7046	3600	1.0598	0.3788	0.9343
0.9485	6.8910	3700	1.0619	0.3312	0.9480
0.9102	7.0764	3800	1.0945	0.3456	0.9804
0.8499	7.2628	3900	1.1076	0.3549	0.9228
0.8106	7.4492	4000	1.0951	0.3419	0.9167
0.8579	7.6356	4100	1.0706	0.3327	0.9749
0.8635	7.8220	4200	1.1565	0.4120	0.9346
0.8444	8.0075	4300	1.0254	0.3385	0.9159
0.7355	8.1938	4400	1.0250	0.3325	0.9072
0.7618	8.3802	4500	1.0682	0.3577	0.9127
0.7531	8.5666	4600	1.0006	0.3206	0.8979
0.7563	8.7530	4700	0.9997	0.3212	0.8957
0.7501	8.9394	4800	1.0075	0.3317	0.9032
0.7056	9.1249	4900	1.0037	0.2974	0.9229
0.6826	9.3113	5000	1.0271	0.2989	0.8882
0.6654	9.4977	5100	1.0321	0.3169	0.8862
0.6754	9.6841	5200	0.9978	0.2944	0.8738
0.6553	9.8705	5300	1.0132	0.3110	0.8893
0.6753	10.0559	5400	0.9893	0.2966	0.8859
0.6155	10.2423	5500	1.0846	0.3168	0.8874
0.6251	10.4287	5600	0.9949	0.3005	0.8828
0.6091	10.6151	5700	1.0041	0.2948	0.8805
0.6384	10.8015	5800	0.9808	0.3000	0.8831
0.626	10.9879	5900	1.0686	0.3335	0.8965
0.5875	11.1733	6000	1.0360	0.3006	0.8880
0.5614	11.3597	6100	1.0932	0.3177	0.8948
0.5736	11.5461	6200	1.0129	0.2963	0.8779
0.5448	11.7325	6300	1.0100	0.2968	0.8727
0.5637	11.9189	6400	1.0889	0.3037	0.9863
0.54	12.1044	6500	0.9968	0.3057	0.8769
0.5235	12.2908	6600	0.9887	0.2921	0.8667
0.5097	12.4772	6700	0.9957	0.3077	0.8841
0.4934	12.6636	6800	1.0222	0.2951	0.8740
0.5302	12.8500	6900	0.9966	0.3023	0.8727
0.4885	13.0354	7000	1.0340	0.2882	0.8657
0.467	13.2218	7100	1.0304	0.2894	0.8620
0.4771	13.4082	7200	0.9793	0.2794	0.8788
0.4577	13.5946	7300	1.0244	0.2904	0.8622
0.4654	13.7810	7400	1.0196	0.2927	0.8703
0.4663	13.9674	7500	1.0139	0.2795	0.8585
0.4335	14.1528	7600	1.0385	0.2799	0.8613
0.4012	14.3392	7700	1.0115	0.2849	0.8758
0.4507	14.5256	7800	0.9969	0.2835	0.8626
0.4558	14.7120	7900	1.0308	0.2897	0.8659
0.4293	14.8984	8000	1.0390	0.2858	0.8608
0.4586	15.0839	8100	1.0493	0.2890	0.9158
0.4009	15.2703	8200	1.0469	0.2914	0.8598
0.3904	15.4567	8300	1.0786	0.2937	0.8703
0.396	15.6431	8400	1.1096	0.2961	0.8705
0.3706	15.8295	8500	1.0184	0.2945	0.875
0.4348	16.0149	8600	1.0295	0.2826	0.8649
0.3488	16.2013	8700	1.0228	0.2874	0.8738
0.3503	16.3877	8800	1.0451	0.2790	0.8547
0.3749	16.5741	8900	1.0531	0.2838	0.8579
0.3409	16.7605	9000	1.0078	0.2846	0.8627
0.3743	16.9469	9100	1.0457	0.2826	0.8726
0.3218	17.1323	9200	1.1266	0.2860	0.8592
0.3168	17.3187	9300	1.0841	0.2868	0.8760
0.3328	17.5051	9400	1.0564	0.2846	0.8762
0.3293	17.6915	9500	1.0349	0.2811	0.8397
0.3189	17.8779	9600	1.0325	0.2773	0.8554
0.3382	18.0634	9700	1.1433	0.2761	0.8425
0.2802	18.2498	9800	1.1109	0.2821	0.8526
0.3114	18.4362	9900	1.0888	0.2846	0.8488
0.2919	18.6226	10000	1.0904	0.2840	0.8451
0.3049	18.8089	10100	1.1035	0.2848	0.8546
0.3129	18.9953	10200	1.1737	0.2825	0.8553
0.2868	19.1808	10300	1.0775	0.2818	0.8528
0.2777	19.3672	10400	1.1083	0.2904	0.8593
0.29	19.5536	10500	1.1036	0.2853	0.8638
0.2948	19.7400	10600	1.1181	0.2834	0.8493
0.3025	19.9264	10700	1.1364	0.2874	0.8691
0.2742	20.1118	10800	1.1488	0.2792	0.8434
0.264	20.2982	10900	1.1320	0.2729	0.8396
0.2652	20.4846	11000	1.1178	0.2773	0.8376
0.2625	20.6710	11100	1.1140	0.2764	0.8345
0.2555	20.8574	11200	1.0976	0.2782	0.8407
0.2827	21.0429	11300	1.1076	0.2818	0.8487
0.2372	21.2293	11400	1.1379	0.2763	0.8480
0.2506	21.4157	11500	1.1521	0.2703	0.8361
0.244	21.6021	11600	1.1719	0.2790	0.8459
0.2477	21.7884	11700	1.1591	0.2749	0.8526
0.2322	21.9748	11800	1.1166	0.2706	0.8373
0.2441	22.1603	11900	1.1119	0.2711	0.8497
0.2151	22.3467	12000	1.1887	0.2791	0.8441
0.2297	22.5331	12100	1.1586	0.2791	0.8493
0.207	22.7195	12200	1.1546	0.2833	0.8496
0.2133	22.9059	12300	1.1111	0.2849	0.8508
0.2245	23.0913	12400	1.1927	0.2806	0.8579
0.2048	23.2777	12500	1.1808	0.2769	0.8400
0.2061	23.4641	12600	1.1464	0.2747	0.8599
0.1879	23.6505	12700	1.1850	0.2737	0.8369
0.2131	23.8369	12800	1.2191	0.2707	0.8405
0.1956	24.0224	12900	1.2319	0.2704	0.8430
0.1903	24.2088	13000	1.2206	0.2691	0.8437
0.1716	24.3952	13100	1.2839	0.2757	0.8461
0.1973	24.5815	13200	1.2227	0.2703	0.8451
0.1933	24.7679	13300	1.2229	0.2746	0.8430
0.1845	24.9543	13400	1.2350	0.2712	0.8390
0.1697	25.1398	13500	1.2309	0.2684	0.8387
0.1763	25.3262	13600	1.1853	0.2668	0.8322
0.1809	25.5126	13700	1.2074	0.2666	0.8379
0.1696	25.6990	13800	1.2031	0.2657	0.8368
0.154	25.8854	13900	1.2128	0.2656	0.8333
0.1946	26.0708	14000	1.2620	0.2685	0.8361
0.1546	26.2572	14100	1.2554	0.2706	0.8356
0.15	26.4436	14200	1.2318	0.2653	0.8374
0.1549	26.6300	14300	1.2386	0.2645	0.8285
0.1679	26.8164	14400	1.2317	0.2665	0.8314
0.1921	27.0019	14500	1.2301	0.2650	0.8369
0.1552	27.1883	14600	1.2628	0.2679	0.8383
0.1492	27.3747	14700	1.2523	0.2666	0.8301
0.146	27.5610	14800	1.2497	0.2669	0.8263
0.1444	27.7474	14900	1.2734	0.2645	0.8416
0.1463	27.9338	15000	1.2638	0.2659	0.8317
0.148	28.1193	15100	1.2823	0.2643	0.8340
0.1473	28.3057	15200	1.2645	0.2649	0.8285
0.1328	28.4921	15300	1.2629	0.2629	0.8298
0.132	28.6785	15400	1.2607	0.2634	0.8253
0.1434	28.8649	15500	1.2769	0.2644	0.8280
0.1392	29.0503	15600	1.2883	0.2649	0.8272
0.1318	29.2367	15700	1.2793	0.2649	0.8264
0.1301	29.4231	15800	1.2833	0.2639	0.8258
0.1297	29.6095	15900	1.2851	0.2638	0.8269
0.1315	29.7959	16000	1.2768	0.2626	0.8260
0.1279	29.9823	16100	1.2753	0.2625	0.8266

Framework versions

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