wav2vec2-age-gender-balancedset

This model is a fine-tuned version of audeering/wav2vec2-large-robust-6-ft-age-gender on the arrow dataset. It achieves the following results on the evaluation set:

• Loss: 2.4630
• Accuracy: 0.7529
• F1 Score: 0.7799
• Mse: 0.7442
• Mae: 0.3779
• Mae^m: 0.3325

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: 9
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 18
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Score	Mse	Mae	Mae^m
1.7071	0.6557	100	1.6599	0.3120	0.2225	3.2857	1.2857	1.2279
1.28	1.3115	200	1.2170	0.5102	0.3996	1.3032	0.7201	0.7433
1.2241	1.9672	300	1.1266	0.5190	0.4383	1.3703	0.7172	0.8739
1.0071	2.6230	400	1.1116	0.5743	0.5019	1.1370	0.6006	0.7637
0.8658	3.2787	500	1.1158	0.6414	0.5407	1.4548	0.6035	0.6665
0.6653	3.9344	600	0.9547	0.6764	0.6021	0.8863	0.4723	0.5315
0.7106	4.5902	700	1.0466	0.6706	0.5908	0.9038	0.4840	0.6695
0.5326	5.2459	800	1.2520	0.6560	0.6961	0.7813	0.4723	0.4067
0.4258	5.9016	900	1.1431	0.6939	0.6080	0.8601	0.4519	0.5134
0.452	6.5574	1000	1.0243	0.7114	0.7066	0.8921	0.4431	0.3818
0.2399	7.2131	1100	1.2856	0.7114	0.7095	0.7901	0.4227	0.3683
0.1564	7.8689	1200	1.5014	0.6676	0.7050	1.1924	0.5394	0.4640
0.3962	8.5246	1300	1.3338	0.7230	0.7545	0.7551	0.4111	0.3599
0.1475	9.1803	1400	1.7256	0.6880	0.7230	0.8688	0.4665	0.4102
0.2757	9.8361	1500	1.6456	0.7085	0.7053	0.6501	0.3994	0.3469
0.1245	10.4918	1600	1.9034	0.7055	0.7464	0.8980	0.4490	0.3869
0.0902	11.1475	1700	1.8943	0.7289	0.7195	0.8163	0.4140	0.3626
0.1188	11.8033	1800	2.0529	0.7376	0.7657	0.7172	0.3848	0.3405
0.0573	12.4590	1900	2.0553	0.7172	0.7512	0.8484	0.4286	0.3666
0.1123	13.1148	2000	2.1915	0.7172	0.7538	0.7405	0.4082	0.3548
0.1463	13.7705	2100	2.0914	0.7259	0.7600	0.6589	0.3790	0.3272
0.3641	14.4262	2200	2.5501	0.6997	0.7388	0.8921	0.4490	0.3839
0.0495	15.0820	2300	2.5900	0.7026	0.6986	0.9038	0.4606	0.3989
0.0177	15.7377	2400	2.2336	0.7201	0.7564	0.8921	0.4373	0.3757
0.0291	16.3934	2500	2.6949	0.7347	0.7692	0.7405	0.3907	0.3391
0.0799	17.0492	2600	2.5497	0.7201	0.7479	0.7172	0.4023	0.3496
0.0982	17.7049	2700	2.4087	0.7464	0.7771	0.6822	0.3732	0.3203
0.072	18.3607	2800	2.2699	0.7289	0.7658	0.7172	0.3965	0.3402
0.1227	19.0164	2900	2.3906	0.7405	0.7758	0.6706	0.3732	0.3188
0.0075	19.6721	3000	2.3322	0.7376	0.7717	0.6472	0.3732	0.3210
0.0019	20.3279	3100	2.4514	0.7434	0.7779	0.6414	0.3673	0.3133
0.0295	20.9836	3200	2.4432	0.7493	0.7813	0.6735	0.3703	0.3170
0.0012	21.6393	3300	2.4851	0.7522	0.7859	0.5831	0.3440	0.2955
0.02	22.2951	3400	2.9030	0.7259	0.7609	0.7347	0.3965	0.3403
0.0341	22.9508	3500	2.6862	0.7347	0.7690	0.7464	0.3965	0.3412
0.0136	23.6066	3600	2.6282	0.7347	0.7685	0.7318	0.3936	0.3415
0.0252	24.2623	3700	2.7268	0.7376	0.7729	0.6297	0.3673	0.3129
0.0403	24.9180	3800	2.5494	0.7434	0.7753	0.6676	0.3703	0.3213
0.0013	25.5738	3900	2.4882	0.7580	0.7890	0.6735	0.3586	0.3083
0.0113	26.2295	4000	2.5213	0.7638	0.7937	0.5627	0.3294	0.2781
0.0371	26.8852	4100	2.6017	0.7638	0.7962	0.5394	0.3294	0.2785
0.0014	27.5410	4200	2.5145	0.7755	0.8044	0.5015	0.3090	0.2616
0.0107	28.1967	4300	2.4742	0.7726	0.8017	0.5131	0.3149	0.2717
0.0002	28.8525	4400	2.4811	0.7609	0.7915	0.5335	0.3294	0.2830
0.0001	29.5082	4500	2.4911	0.7609	0.7921	0.5481	0.3324	0.2842

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1