ssc-meh-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: 1.8542
• Cer: 0.3340
• Wer: 0.8345

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
5.7202	0.1905	100	2.9202	1.0	1.0
2.9539	0.3810	200	2.7751	1.0	1.0
2.9096	0.5714	300	2.7931	1.0	1.0
2.9036	0.7619	400	2.7883	1.0	1.0
2.8563	0.9524	500	2.9839	1.0	1.0
2.7445	1.1429	600	2.5705	1.0	1.0
2.6738	1.3333	700	2.4914	0.9969	0.9992
2.5821	1.5238	800	2.2302	0.9231	0.9753
2.4371	1.7143	900	2.0914	0.8529	1.0
2.2815	1.9048	1000	1.8929	0.7330	0.9997
2.1294	2.0952	1100	1.8964	0.6461	0.9673
1.9156	2.2857	1200	1.8122	0.7197	0.9931
1.8408	2.4762	1300	1.7276	0.5993	0.9602
1.7506	2.6667	1400	1.6561	0.6534	0.9657
1.7086	2.8571	1500	1.7104	0.4768	0.9349
1.5834	3.0476	1600	1.4368	0.4460	0.9084
1.5731	3.2381	1700	1.4594	0.4292	0.9379
1.5743	3.4286	1800	1.4713	0.4183	0.9731
1.4766	3.6190	1900	1.3967	0.4334	0.9114
1.4799	3.8095	2000	1.5286	0.4260	0.9550
1.4973	4.0	2100	1.7988	0.4115	1.1627
1.3498	4.1905	2200	1.4739	0.4768	0.9209
1.3642	4.3810	2300	1.3685	0.4241	0.9290
1.361	4.5714	2400	1.3578	0.4705	0.9097
1.3747	4.7619	2500	1.4424	0.4560	0.9060
1.3728	4.9524	2600	1.5908	0.3978	0.9865
1.2524	5.1429	2700	1.3019	0.3814	0.9119
1.2532	5.3333	2800	1.3969	0.3960	0.9056
1.2527	5.5238	2900	1.2484	0.3659	0.9150
1.2347	5.7143	3000	1.5303	0.4045	1.0358
1.2531	5.9048	3100	1.3037	0.3827	0.8702
1.2506	6.0952	3200	1.4484	0.3815	0.9621
1.1295	6.2857	3300	1.4467	0.3843	0.9590
1.1507	6.4762	3400	1.3074	0.4033	0.8888
1.146	6.6667	3500	1.6379	0.4109	1.0024
1.1812	6.8571	3600	1.3674	0.4137	0.9033
1.063	7.0476	3700	1.2695	0.3719	0.9200
1.0615	7.2381	3800	1.2816	0.3645	0.9137
1.0779	7.4286	3900	1.3308	0.3772	0.9202
1.059	7.6190	4000	1.1988	0.3983	0.8730
1.0706	7.8095	4100	1.2735	0.3771	0.8901
1.1306	8.0	4200	1.3397	0.4057	0.8777
0.965	8.1905	4300	1.4966	0.3934	1.0157
1.0081	8.3810	4400	1.2140	0.3899	0.8674
1.0052	8.5714	4500	1.2490	0.3873	0.8774
1.0262	8.7619	4600	1.2793	0.3705	0.8633
0.9791	8.9524	4700	1.1578	0.3542	0.8501
0.8975	9.1429	4800	1.1994	0.3646	0.8731
0.9043	9.3333	4900	1.1890	0.3426	0.8523
0.9641	9.5238	5000	1.2537	0.3604	0.9406
0.9526	9.7143	5100	1.2081	0.3549	0.8456
0.905	9.9048	5200	1.2440	0.3614	0.9021
0.9099	10.0952	5300	1.3497	0.3703	0.9114
0.8435	10.2857	5400	1.1731	0.3861	0.8658
0.8914	10.4762	5500	1.3744	0.3745	0.9399
0.879	10.6667	5600	1.2298	0.3559	0.8850
0.8754	10.8571	5700	1.3239	0.3508	0.9467
0.8348	11.0476	5800	1.2166	0.3551	0.8572
0.781	11.2381	5900	1.2808	0.3511	0.9238
0.8081	11.4286	6000	1.2105	0.3496	0.8970
0.8023	11.6190	6100	1.1924	0.3434	0.8323
0.8236	11.8095	6200	1.2141	0.3453	0.9049
0.892	12.0	6300	1.2798	0.3720	0.9378
0.7549	12.1905	6400	1.1700	0.3318	0.8809
0.7428	12.3810	6500	1.3049	0.3628	0.8782
0.765	12.5714	6600	1.2451	0.3449	0.9348
0.7747	12.7619	6700	1.3216	0.3714	0.8654
0.7448	12.9524	6800	1.3266	0.3633	0.8783
0.6694	13.1429	6900	1.2095	0.3411	0.8247
0.6949	13.3333	7000	1.3466	0.3589	0.8935
0.721	13.5238	7100	1.2278	0.3478	0.8737
0.7298	13.7143	7200	1.2689	0.3533	0.9153
0.6928	13.9048	7300	1.2013	0.3433	0.8368
0.7061	14.0952	7400	1.3039	0.3781	0.8541
0.6477	14.2857	7500	1.2427	0.3407	0.8845
0.6752	14.4762	7600	1.3106	0.3536	0.8776
0.6498	14.6667	7700	1.2447	0.3412	0.8615
0.6626	14.8571	7800	1.3016	0.3621	0.8733
0.5813	15.0476	7900	1.2265	0.3362	0.8301
0.6195	15.2381	8000	1.2933	0.3445	0.8307
0.5909	15.4286	8100	1.2376	0.3351	0.8288
0.6048	15.6190	8200	1.2648	0.3374	0.8197
0.6235	15.8095	8300	1.2515	0.3410	0.8602
0.6721	16.0	8400	1.2712	0.3422	0.8459
0.5156	16.1905	8500	1.3692	0.3527	0.8760
0.5312	16.3810	8600	1.3574	0.3438	0.8598
0.5859	16.5714	8700	1.2556	0.3398	0.8362
0.5454	16.7619	8800	1.3780	0.3574	0.8749
0.5552	16.9524	8900	1.2868	0.3461	0.8953
0.484	17.1429	9000	1.2784	0.3434	0.8335
0.5364	17.3333	9100	1.2629	0.3339	0.8294
0.5188	17.5238	9200	1.3447	0.3329	0.8361
0.4979	17.7143	9300	1.3755	0.3332	0.8532
0.5313	17.9048	9400	1.2748	0.3397	0.8486
0.5102	18.0952	9500	1.3303	0.3363	0.8214
0.4997	18.2857	9600	1.3119	0.3453	0.8571
0.475	18.4762	9700	1.4079	0.3446	0.8573
0.4989	18.6667	9800	1.3263	0.3335	0.8518
0.4915	18.8571	9900	1.3138	0.3410	0.8241
0.4469	19.0476	10000	1.3820	0.3309	0.8491
0.4391	19.2381	10100	1.3812	0.3344	0.8371
0.4481	19.4286	10200	1.3321	0.3434	0.8303
0.4482	19.6190	10300	1.3548	0.3356	0.8298
0.4592	19.8095	10400	1.3296	0.3405	0.8610
0.4886	20.0	10500	1.3636	0.3515	0.8321
0.4023	20.1905	10600	1.4382	0.3514	0.8467
0.4332	20.3810	10700	1.2829	0.3319	0.8508
0.4154	20.5714	10800	1.3922	0.3423	0.8313
0.4276	20.7619	10900	1.4043	0.3459	0.8460
0.4316	20.9524	11000	1.4012	0.3346	0.8272
0.351	21.1429	11100	1.4923	0.3404	0.8368
0.3941	21.3333	11200	1.4509	0.3438	0.8382
0.3883	21.5238	11300	1.4189	0.3359	0.8258
0.4208	21.7143	11400	1.4527	0.3411	0.8344
0.3843	21.9048	11500	1.5000	0.3474	0.8349
0.42	22.0952	11600	1.6168	0.3509	0.8666
0.3638	22.2857	11700	1.5645	0.3518	0.8581
0.3763	22.4762	11800	1.4347	0.3441	0.8419
0.3637	22.6667	11900	1.6041	0.3466	0.8635
0.3717	22.8571	12000	1.5876	0.3400	0.8466
0.3398	23.0476	12100	1.5634	0.3378	0.8281
0.347	23.2381	12200	1.4949	0.3316	0.8112
0.3493	23.4286	12300	1.5127	0.3427	0.8347
0.338	23.6190	12400	1.5340	0.3458	0.8423
0.3456	23.8095	12500	1.5608	0.3492	0.8532
0.3829	24.0	12600	1.5481	0.3391	0.8569
0.331	24.1905	12700	1.5683	0.3323	0.8539
0.305	24.3810	12800	1.6475	0.3298	0.8240
0.3244	24.5714	12900	1.5566	0.3375	0.8227
0.3397	24.7619	13000	1.5676	0.3368	0.8393
0.3221	24.9524	13100	1.5839	0.3376	0.8391
0.2943	25.1429	13200	1.7326	0.3316	0.8413
0.3197	25.3333	13300	1.6815	0.3329	0.8492
0.3011	25.5238	13400	1.6816	0.3359	0.8492
0.3152	25.7143	13500	1.6790	0.3391	0.8683
0.3088	25.9048	13600	1.6628	0.3428	0.8819
0.336	26.0952	13700	1.7037	0.3340	0.8444
0.2778	26.2857	13800	1.7835	0.3479	0.8627
0.2775	26.4762	13900	1.7454	0.3359	0.8630
0.286	26.6667	14000	1.7595	0.3359	0.8395
0.2724	26.8571	14100	1.7468	0.3387	0.8408
0.2663	27.0476	14200	1.6998	0.3352	0.8452
0.2941	27.2381	14300	1.8233	0.3367	0.8526
0.2888	27.4286	14400	1.7692	0.3353	0.8241
0.2511	27.6190	14500	1.8221	0.3371	0.8425
0.2623	27.8095	14600	1.8068	0.3400	0.8403
0.2851	28.0	14700	1.8250	0.3398	0.8417
0.2557	28.1905	14800	1.8385	0.3405	0.8464
0.2844	28.3810	14900	1.8351	0.3359	0.8456
0.25	28.5714	15000	1.8201	0.3326	0.8261
0.267	28.7619	15100	1.8391	0.3376	0.8375
0.2409	28.9524	15200	1.8450	0.3371	0.8430
0.2263	29.1429	15300	1.8316	0.3356	0.8336
0.2579	29.3333	15400	1.8472	0.3338	0.8391
0.2424	29.5238	15500	1.8536	0.3328	0.8357
0.2612	29.7143	15600	1.8561	0.3341	0.8339
0.2644	29.9048	15700	1.8542	0.3340	0.8345

Framework versions

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