wav2vec2-turkish-tr-voice

This model is a fine-tuned version of facebook/wav2vec2-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0239

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: 0.0001
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 8
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.479	0.0765	1000	0.2672
0.3629	0.1531	2000	0.1266
0.2485	0.2296	3000	0.1020
0.2082	0.3062	4000	0.0916
0.1913	0.3827	5000	0.0737
0.1757	0.4593	6000	0.0741
0.1638	0.5358	7000	0.0686
0.1567	0.6124	8000	0.0636
0.153	0.6889	9000	0.0608
0.1488	0.7655	10000	0.0583
0.1374	0.8420	11000	0.0536
0.1357	0.9186	12000	0.0511
0.1393	0.9951	13000	0.0525
0.1314	1.0716	14000	0.0491
0.1194	1.1482	15000	0.0498
0.1277	1.2247	16000	0.0484
0.1173	1.3013	17000	0.0443
0.1214	1.3778	18000	0.0443
0.1136	1.4544	19000	0.0453
0.1107	1.5309	20000	0.0444
0.1155	1.6075	21000	0.0419
0.1102	1.6840	22000	0.0406
0.107	1.7606	23000	0.0414
0.104	1.8371	24000	0.0397
0.1095	1.9137	25000	0.0405
0.1004	1.9902	26000	0.0370
0.1038	2.0667	27000	0.0394
0.097	2.1433	28000	0.0422
0.0983	2.2198	29000	0.0369
0.0953	2.2964	30000	0.0405
0.0958	2.3729	31000	0.0362
0.0934	2.4495	32000	0.0376
0.0939	2.5260	33000	0.0348
0.0909	2.6026	34000	0.0356
0.0927	2.6791	35000	0.0332
0.0915	2.7557	36000	0.0324
0.0933	2.8322	37000	0.0354
0.0905	2.9088	38000	0.0328
0.0905	2.9853	39000	0.0351
0.0869	3.0618	40000	0.0303
0.0869	3.1384	41000	0.0351
0.0877	3.2149	42000	0.0339
0.0859	3.2915	43000	0.0342
0.0922	3.3680	44000	0.0331
0.0848	3.4446	45000	0.0350
0.0836	3.5211	46000	0.0307
0.0887	3.5977	47000	0.0317
0.088	3.6742	48000	0.0298
0.083	3.7508	49000	0.0325
0.0824	3.8273	50000	0.0318
0.0811	3.9039	51000	0.0315
0.0797	3.9804	52000	0.0289
0.0794	4.0570	53000	0.0330
0.0844	4.1335	54000	0.0325
0.0784	4.2100	55000	0.0330
0.0765	4.2866	56000	0.0295
0.0793	4.3631	57000	0.0293
0.0775	4.4397	58000	0.0302
0.076	4.5162	59000	0.0283
0.074	4.5928	60000	0.0277
0.0766	4.6693	61000	0.0296
0.074	4.7459	62000	0.0274
0.0744	4.8224	63000	0.0276
0.0784	4.8990	64000	0.0295
0.0763	4.9755	65000	0.0277
0.0722	5.0521	66000	0.0290
0.0713	5.1286	67000	0.0277
0.0708	5.2051	68000	0.0308
0.0714	5.2817	69000	0.0300
0.0742	5.3582	70000	0.0273
0.0717	5.4348	71000	0.0261
0.0699	5.5113	72000	0.0277
0.0695	5.5879	73000	0.0275
0.0686	5.6644	74000	0.0267
0.0702	5.7410	75000	0.0272
0.0672	5.8175	76000	0.0269
0.0724	5.8941	77000	0.0274
0.0696	5.9706	78000	0.0246
0.07	6.0472	79000	0.0325
0.0903	6.1237	80000	0.0276
0.0693	6.2002	81000	0.0250
0.0714	6.2768	82000	0.0255
0.0655	6.3533	83000	0.0258
0.0652	6.4299	84000	0.0270
0.0709	6.5064	85000	0.0253
0.0666	6.5830	86000	0.0253
0.0678	6.6595	87000	0.0257
0.0692	6.7361	88000	0.0236
0.0657	6.8126	89000	0.0287
0.0657	6.8892	90000	0.0240
0.0646	6.9657	91000	0.0245
0.0616	7.0423	92000	0.0254
0.0653	7.1188	93000	0.0291
0.0653	7.1953	94000	0.0253
0.0617	7.2719	95000	0.0239
0.0616	7.3484	96000	0.0245
0.0661	7.4250	97000	0.0237
0.0629	7.5015	98000	0.0239

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1