wav2vec2-turkish-300m-2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_16_1 dataset. It achieves the following results on the evaluation set:

Loss: 0.3293
Wer: 0.2786

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: 16
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 0.1
num_epochs: 20
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.234	0.1461	400	0.7428	0.7198
0.6888	0.2922	800	0.7178	0.7417
0.5978	0.4383	1200	0.5479	0.6138
0.5608	0.5844	1600	0.5362	0.5827
0.52	0.7305	2000	0.6510	0.6688
0.5019	0.8766	2400	0.5023	0.5676
0.4791	1.0226	2800	0.4218	0.5065
0.4217	1.1687	3200	0.4133	0.4860
0.416	1.3148	3600	0.4329	0.4964
0.4067	1.4609	4000	0.4084	0.4871
0.4046	1.6070	4400	0.4238	0.5113
0.4108	1.7531	4800	0.4063	0.4854
0.4018	1.8992	5200	0.4123	0.4861
0.3878	2.0453	5600	0.3969	0.4763
0.351	2.1914	6000	0.4012	0.4690
0.3454	2.3375	6400	0.4011	0.4741
0.3487	2.4836	6800	0.3908	0.4816
0.3417	2.6297	7200	0.3884	0.4578
0.3408	2.7757	7600	0.4002	0.4674
0.3374	2.9218	8000	0.3914	0.4524
0.3121	3.0679	8400	0.4217	0.4763
0.3025	3.2140	8800	0.3905	0.4598
0.3016	3.3601	9200	0.3831	0.4505
0.2974	3.5062	9600	0.3948	0.4568
0.3032	3.6523	10000	0.3888	0.4512
0.2929	3.7984	10400	0.3900	0.4461
0.2922	3.9445	10800	0.3846	0.4540
0.2745	4.0906	11200	0.3748	0.4374
0.2665	4.2367	11600	0.3993	0.4336
0.2696	4.3828	12000	0.3657	0.4333
0.2622	4.5289	12400	0.3628	0.4341
0.2585	4.6749	12800	0.3802	0.4288
0.2604	4.8210	13200	0.3604	0.4236
0.2575	4.9671	13600	0.3650	0.4322
0.2313	5.1132	14000	0.3521	0.4021
0.2323	5.2593	14400	0.3513	0.4009
0.2351	5.4054	14800	0.3265	0.3963
0.2229	5.5515	15200	0.3523	0.3978
0.234	5.6976	15600	0.3375	0.3931
0.229	5.8437	16000	0.3380	0.3945
0.2313	5.9898	16400	0.3403	0.3957
0.2069	6.1359	16800	0.3522	0.3979
0.2162	6.2820	17200	0.3685	0.4061
0.2144	6.4280	17600	0.3308	0.3878
0.2115	6.5741	18000	0.3530	0.3974
0.2108	6.7202	18400	0.3191	0.3802
0.2107	6.8663	18800	0.3313	0.3818
0.1977	7.0124	19200	0.3454	0.3807
0.1903	7.1585	19600	0.3386	0.3785
0.1924	7.3046	20000	0.3369	0.3841
0.1912	7.4507	20400	0.3385	0.3782
0.1879	7.5968	20800	0.3302	0.3728
0.1903	7.7429	21200	0.3254	0.3636
0.1828	7.8890	21600	0.3499	0.3723
0.1803	8.0351	22000	0.3371	0.3834
0.1711	8.1812	22400	0.3498	0.3879
0.169	8.3272	22800	0.3332	0.3731
0.1657	8.4733	23200	0.3223	0.3665
0.1682	8.6194	23600	0.3386	0.3696
0.1732	8.7655	24000	0.3564	0.3726
0.1723	8.9116	24400	0.3336	0.3685
0.1681	9.0577	24800	0.3328	0.3543
0.1547	9.2038	25200	0.3358	0.3533
0.1572	9.3499	25600	0.3088	0.3563
0.1518	9.4960	26000	0.3219	0.3513
0.1532	9.6421	26400	0.3060	0.3491
0.154	9.7882	26800	0.3091	0.3457
0.1478	9.9343	27200	0.3159	0.3401
0.1499	10.0804	27600	0.3219	0.3485
0.1337	10.2264	28000	0.3109	0.3443
0.1364	10.3725	28400	0.3281	0.3456
0.1329	10.5186	28800	0.3143	0.3408
0.146	10.6647	29200	0.3285	0.3383
0.1403	10.8108	29600	0.3180	0.3387
0.14	10.9569	30000	0.3086	0.3350
0.1258	11.1030	30400	0.3253	0.3345
0.1229	11.2491	30800	0.3236	0.3392
0.1241	11.3952	31200	0.3257	0.3349
0.1224	11.5413	31600	0.3260	0.3287
0.1218	11.6874	32000	0.3153	0.3330
0.1267	11.8335	32400	0.3141	0.3298
0.1246	11.9795	32800	0.3144	0.3281
0.113	12.1256	33200	0.3415	0.3367
0.1121	12.2717	33600	0.3262	0.3294
0.1147	12.4178	34000	0.3378	0.3287
0.114	12.5639	34400	0.3121	0.3240
0.1054	12.7100	34800	0.3288	0.3199
0.1081	12.8561	35200	0.3010	0.3220
0.1137	13.0022	35600	0.3261	0.3229
0.102	13.1483	36000	0.3168	0.3177
0.1	13.2944	36400	0.3224	0.3173
0.1003	13.4405	36800	0.3175	0.3205
0.098	13.5866	37200	0.3021	0.3158
0.0974	13.7327	37600	0.3057	0.3154
0.0952	13.8787	38000	0.3257	0.3155
0.095	14.0248	38400	0.3229	0.3097
0.0902	14.1709	38800	0.3285	0.3152
0.0917	14.3170	39200	0.3279	0.3160
0.0905	14.4631	39600	0.3278	0.3111
0.092	14.6092	40000	0.3209	0.3105
0.0862	14.7553	40400	0.3109	0.3064
0.0912	14.9014	40800	0.3116	0.3056
0.086	15.0475	41200	0.3383	0.3038
0.0832	15.1936	41600	0.3189	0.3018
0.0773	15.3397	42000	0.3150	0.3033
0.0817	15.4858	42400	0.3253	0.3040
0.0775	15.6318	42800	0.3223	0.3030
0.0767	15.7779	43200	0.3225	0.2970
0.0796	15.9240	43600	0.3368	0.3047
0.0763	16.0701	44000	0.3252	0.2971
0.075	16.2162	44400	0.3188	0.3002
0.0744	16.3623	44800	0.3207	0.2947
0.0729	16.5084	45200	0.3214	0.2956
0.0732	16.6545	45600	0.3278	0.2927
0.0694	16.8006	46000	0.3364	0.2924
0.0753	16.9467	46400	0.3263	0.2881
0.072	17.0928	46800	0.3317	0.2909
0.0658	17.2389	47200	0.3376	0.2902
0.07	17.3850	47600	0.3282	0.2902
0.0636	17.5310	48000	0.3321	0.2920
0.0655	17.6771	48400	0.3274	0.2884
0.0623	17.8232	48800	0.3316	0.2890
0.0621	17.9693	49200	0.3209	0.2872
0.0637	18.1154	49600	0.3281	0.2830
0.0616	18.2615	50000	0.3393	0.2852
0.0605	18.4076	50400	0.3371	0.2849
0.0614	18.5537	50800	0.3277	0.2836
0.0571	18.6998	51200	0.3317	0.2821
0.0555	18.8459	51600	0.3364	0.2816
0.061	18.9920	52000	0.3251	0.2797
0.0571	19.1381	52400	0.3343	0.2811
0.0614	19.2841	52800	0.3300	0.2810
0.0534	19.4302	53200	0.3324	0.2795
0.0562	19.5763	53600	0.3322	0.2789
0.0545	19.7224	54000	0.3310	0.2789
0.0596	19.8685	54400	0.3293	0.2786

Framework versions

Transformers 4.40.0
Pytorch 2.2.2+cu121
Datasets 2.17.1
Tokenizers 0.19.1

tgrhn/wav2vec2-turkish-300m-2

wav2vec2-turkish-300m-2

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

Seed this model