ssc-lth-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: 1.5683
• Cer: 0.3085
• Wer: 0.7821

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
5.7582	0.1546	100	3.1629	0.9943	1.0
3.4065	0.3091	200	3.1093	0.9943	1.0
3.3576	0.4637	300	3.0748	0.9943	1.0
3.1832	0.6182	400	3.0296	0.9943	1.0
3.1613	0.7728	500	2.9966	0.9943	1.0
3.1446	0.9274	600	3.0421	0.9943	1.0
3.1082	1.0819	700	3.0424	0.9943	1.0
3.0153	1.2365	800	2.8610	0.9943	1.0
2.8793	1.3910	900	2.8591	0.8887	1.0
2.7603	1.5456	1000	2.5400	0.8918	1.0
2.6694	1.7002	1100	2.4441	0.8234	1.0
2.4981	1.8547	1200	2.1997	0.7659	1.0
2.3687	2.0093	1300	2.0222	0.7077	1.0
2.3179	2.1638	1400	1.9736	0.6633	1.0
2.2323	2.3184	1500	2.0691	0.5410	0.9733
2.1893	2.4730	1600	1.8443	0.5961	0.9948
2.1184	2.6275	1700	1.8428	0.5332	0.9794
2.0585	2.7821	1800	1.7327	0.5545	0.9920
2.0365	2.9366	1900	1.7461	0.5172	0.9742
1.9224	3.0912	2000	1.6829	0.4921	0.9501
1.9396	3.2457	2100	1.6787	0.4800	0.9399
1.9326	3.4003	2200	1.6375	0.5063	0.9644
1.8535	3.5549	2300	1.6068	0.4837	0.9530
1.859	3.7094	2400	1.5812	0.4838	0.9517
1.8666	3.8640	2500	1.6099	0.4670	0.9356
1.8084	4.0185	2600	1.5219	0.4707	0.9381
1.7369	4.1731	2700	1.5108	0.4377	0.9229
1.7619	4.3277	2800	1.4897	0.4434	0.9216
1.7268	4.4822	2900	1.5166	0.4525	0.9377
1.6908	4.6368	3000	1.5185	0.4541	0.9247
1.7478	4.7913	3100	1.4809	0.4382	0.9254
1.7223	4.9459	3200	1.4538	0.4354	0.9231
1.6529	5.1005	3300	1.4845	0.4355	0.9185
1.633	5.2550	3400	1.4235	0.4358	0.9252
1.6486	5.4096	3500	1.3745	0.4156	0.9040
1.6362	5.5641	3600	1.4383	0.4047	0.9683
1.552	5.7187	3700	1.4306	0.4195	0.9150
1.5794	5.8733	3800	1.4064	0.4153	0.9155
1.5882	6.0278	3900	1.4026	0.4055	0.9205
1.4888	6.1824	4000	1.4291	0.4131	0.9315
1.5515	6.3369	4100	1.4047	0.4018	0.9229
1.5229	6.4915	4200	1.3459	0.3977	0.8988
1.5235	6.6461	4300	1.3754	0.3957	0.9233
1.4919	6.8006	4400	1.3355	0.3897	0.9222
1.5215	6.9552	4500	1.3391	0.3882	0.9048
1.4768	7.1097	4600	1.3710	0.3982	0.8937
1.4441	7.2643	4700	1.3814	0.3961	0.9112
1.3782	7.4189	4800	1.3806	0.4105	0.9028
1.4264	7.5734	4900	1.3845	0.3897	0.9491
1.4418	7.7280	5000	1.2996	0.3800	0.8863
1.4022	7.8825	5100	1.3480	0.3957	0.8869
1.4414	8.0371	5200	1.3391	0.3847	0.9246
1.3637	8.1917	5300	1.3786	0.3991	0.8862
1.365	8.3462	5400	1.2898	0.3729	0.8711
1.3352	8.5008	5500	1.2853	0.3774	0.8765
1.3346	8.6553	5600	1.3410	0.3974	0.8972
1.3163	8.8099	5700	1.3299	0.3836	0.9205
1.38	8.9645	5800	1.4417	0.3892	0.9467
1.3366	9.1190	5900	1.4005	0.3906	0.8904
1.287	9.2736	6000	1.3038	0.3713	0.8920
1.2572	9.4281	6100	1.3145	0.3853	0.8755
1.2785	9.5827	6200	1.3774	0.3848	0.8795
1.237	9.7372	6300	1.3252	0.3853	0.8768
1.2779	9.8918	6400	1.2781	0.3737	0.8769
1.2935	10.0464	6500	1.3950	0.3881	0.8870
1.1689	10.2009	6600	1.3417	0.3744	0.8764
1.2232	10.3555	6700	1.3423	0.3755	0.8663
1.2236	10.5100	6800	1.2751	0.3684	0.8837
1.2102	10.6646	6900	1.2827	0.3716	0.8638
1.2025	10.8192	7000	1.3115	0.3707	0.8959
1.2348	10.9737	7100	1.3197	0.3633	0.8900
1.1198	11.1283	7200	1.3162	0.3682	0.8566
1.1254	11.2828	7300	1.3437	0.3709	0.8776
1.1438	11.4374	7400	1.3546	0.3753	0.8628
1.1528	11.5920	7500	1.2441	0.3541	0.8558
1.1301	11.7465	7600	1.2736	0.3697	0.8635
1.1229	11.9011	7700	1.2383	0.3597	0.8637
1.1234	12.0556	7800	1.2476	0.3583	0.8518
1.0714	12.2102	7900	1.2173	0.3515	0.8411
1.0748	12.3648	8000	1.1947	0.3478	0.8346
1.0588	12.5193	8100	1.3123	0.3533	0.8537
1.0727	12.6739	8200	1.2479	0.3562	0.8472
1.0688	12.8284	8300	1.2803	0.3600	0.8502
1.1059	12.9830	8400	1.2793	0.3529	0.8476
1.0258	13.1376	8500	1.2690	0.3636	0.8521
1.012	13.2921	8600	1.2987	0.3478	0.8704
0.9649	13.4467	8700	1.4050	0.3718	0.8624
0.9875	13.6012	8800	1.2196	0.3451	0.8436
1.0292	13.7558	8900	1.2596	0.3529	0.8320
1.0342	13.9104	9000	1.2486	0.3464	0.8497
1.0011	14.0649	9100	1.2429	0.3533	0.8423
0.9401	14.2195	9200	1.3330	0.3603	0.8569
0.9597	14.3740	9300	1.2858	0.3636	0.8501
0.9457	14.5286	9400	1.1997	0.3444	0.8472
0.9661	14.6832	9500	1.1897	0.3401	0.8425
0.9555	14.8377	9600	1.2808	0.3527	0.8334
0.9833	14.9923	9700	1.2827	0.3486	0.8392
0.8874	15.1468	9800	1.4258	0.3675	0.8566
0.8955	15.3014	9900	1.2527	0.3462	0.8229
0.8936	15.4560	10000	1.2336	0.3363	0.8394
0.8948	15.6105	10100	1.2048	0.3366	0.8244
0.9023	15.7651	10200	1.3404	0.3410	0.8237
0.9056	15.9196	10300	1.3217	0.3543	0.8379
0.904	16.0742	10400	1.3742	0.3638	0.8376
0.7977	16.2287	10500	1.2720	0.3305	0.8096
0.8473	16.3833	10600	1.3408	0.3376	0.8546
0.8579	16.5379	10700	1.2352	0.3421	0.8193
0.8524	16.6924	10800	1.3313	0.3500	0.8385
0.8388	16.8470	10900	1.2586	0.3315	0.8296
0.8291	17.0015	11000	1.2370	0.3517	0.8368
0.7695	17.1561	11100	1.2832	0.3339	0.8271
0.7844	17.3107	11200	1.2764	0.3453	0.8468
0.7764	17.4652	11300	1.2639	0.3428	0.8409
0.8082	17.6198	11400	1.2235	0.3359	0.8198
0.7982	17.7743	11500	1.2658	0.3352	0.8140
0.7763	17.9289	11600	1.2210	0.3287	0.8186
0.7947	18.0835	11700	1.2735	0.3292	0.7998
0.7261	18.2380	11800	1.2937	0.3400	0.8121
0.745	18.3926	11900	1.2504	0.3385	0.8136
0.7626	18.5471	12000	1.2307	0.3260	0.8023
0.725	18.7017	12100	1.2764	0.3369	0.8021
0.7459	18.8563	12200	1.2189	0.3309	0.8020
0.7395	19.0108	12300	1.2120	0.3197	0.7927
0.689	19.1654	12400	1.2412	0.3241	0.7974
0.6776	19.3199	12500	1.1921	0.3202	0.7936
0.6629	19.4745	12600	1.2220	0.3220	0.7943
0.7292	19.6291	12700	1.2844	0.3186	0.7953
0.7135	19.7836	12800	1.2520	0.3211	0.7932
0.7042	19.9382	12900	1.2142	0.3127	0.7783
0.6766	20.0927	13000	1.2213	0.3176	0.7884
0.6471	20.2473	13100	1.3010	0.3232	0.8033
0.6842	20.4019	13200	1.3410	0.3273	0.8023
0.6308	20.5564	13300	1.2506	0.3159	0.7932
0.6551	20.7110	13400	1.3087	0.3197	0.7979
0.6413	20.8655	13500	1.2666	0.3168	0.7988
0.6279	21.0201	13600	1.3987	0.3141	0.7816
0.6173	21.1747	13700	1.3371	0.3125	0.7974
0.5904	21.3292	13800	1.3081	0.3274	0.8015
0.6126	21.4838	13900	1.3683	0.3247	0.8083
0.6055	21.6383	14000	1.3113	0.3157	0.7977
0.6225	21.7929	14100	1.2706	0.3110	0.7865
0.6007	21.9474	14200	1.3128	0.3264	0.7931
0.5837	22.1020	14300	1.3846	0.3198	0.7923
0.5531	22.2566	14400	1.3394	0.3265	0.8053
0.5748	22.4111	14500	1.3888	0.3220	0.7889
0.555	22.5657	14600	1.3188	0.3242	0.7950
0.5632	22.7202	14700	1.3423	0.3216	0.7884
0.5717	22.8748	14800	1.3724	0.3133	0.8009
0.5513	23.0294	14900	1.3964	0.3183	0.7924
0.5324	23.1839	15000	1.4374	0.3187	0.8119
0.5538	23.3385	15100	1.3371	0.3176	0.7888
0.5098	23.4930	15200	1.3456	0.3139	0.7812
0.5437	23.6476	15300	1.3673	0.3183	0.7910
0.5621	23.8022	15400	1.3920	0.3204	0.7893
0.5265	23.9567	15500	1.3835	0.3211	0.7995
0.4964	24.1113	15600	1.3911	0.3153	0.7877
0.489	24.2658	15700	1.4370	0.3224	0.8030
0.4907	24.4204	15800	1.4016	0.3268	0.7973
0.49	24.5750	15900	1.4152	0.3175	0.7901
0.5104	24.7295	16000	1.4569	0.3200	0.7846
0.5075	24.8841	16100	1.4782	0.3175	0.7957
0.4939	25.0386	16200	1.4216	0.3155	0.7935
0.4501	25.1932	16300	1.4857	0.3234	0.7995
0.4498	25.3478	16400	1.4586	0.3161	0.7969
0.4521	25.5023	16500	1.5067	0.3195	0.7895
0.4624	25.6569	16600	1.4472	0.3151	0.7927
0.4881	25.8114	16700	1.4872	0.3175	0.7999
0.4802	25.9660	16800	1.4425	0.3118	0.7969
0.4441	26.1206	16900	1.5231	0.3177	0.8007
0.4641	26.2751	17000	1.4957	0.3152	0.7935
0.4417	26.4297	17100	1.5059	0.3151	0.7932
0.4492	26.5842	17200	1.4759	0.3136	0.7847
0.4443	26.7388	17300	1.5226	0.3121	0.7882
0.4635	26.8934	17400	1.4899	0.3090	0.7768
0.4432	27.0479	17500	1.5173	0.3156	0.7818
0.4349	27.2025	17600	1.5088	0.3122	0.7817
0.4235	27.3570	17700	1.5343	0.3149	0.8004
0.4174	27.5116	17800	1.5429	0.3090	0.7906
0.4293	27.6662	17900	1.5188	0.3122	0.7838
0.4164	27.8207	18000	1.5407	0.3090	0.7850
0.4202	27.9753	18100	1.5115	0.3092	0.7797
0.393	28.1298	18200	1.5165	0.3082	0.7799
0.3997	28.2844	18300	1.5150	0.3093	0.7825
0.4368	28.4389	18400	1.5381	0.3092	0.7765
0.372	28.5935	18500	1.5266	0.3096	0.7876
0.4138	28.7481	18600	1.5435	0.3085	0.7823
0.4082	28.9026	18700	1.5375	0.3079	0.7825
0.4009	29.0572	18800	1.5643	0.3098	0.7856
0.3883	29.2117	18900	1.5731	0.3063	0.7829
0.3836	29.3663	19000	1.5718	0.3073	0.7820
0.3939	29.5209	19100	1.5822	0.3079	0.7827
0.3915	29.6754	19200	1.5766	0.3087	0.7829
0.3958	29.8300	19300	1.5688	0.3086	0.7827
0.4069	29.9845	19400	1.5683	0.3085	0.7821

Framework versions

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