jialicheng/unlearn_speech_commands_wav2vec2-base_random_label_8_42

superb_ks_42

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

• Loss: 0.1831
• Accuracy: 0.9826
• Test Accuracy: 0.9826
• Df Accuracy: 0.9812
• Unlearn Overall Accuracy: 0.5007
• Unlearn Time: 4465.4075

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: 5e-05
• train_batch_size: 32
• 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_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Overall Accuracy	Unlearn Overall Accuracy	Time
0.4658	1.0	1597	0.1831	0.9812	0.5007	0.5007	-1
0.3682	2.0	3194	0.1797	0.9804	0.4991	0.4991	-1
0.375	3.0	4791	0.2095	0.9763	0.5003	0.5003	-1
0.375	4.0	6388	0.1657	0.9802	0.5005	0.5005	-1
0.371	5.0	7985	0.1792	0.9809	0.4990	0.4990	-1
0.3501	6.0	9582	0.1797	0.9821	0.4982	0.4982	-1
0.3452	7.0	11179	0.1898	0.9819	0.4991	0.4991	-1
0.3397	8.0	12776	0.1645	0.9831	0.4998	0.4998	-1
0.3362	9.0	14373	0.1814	0.9809	0.5000	0.5000	-1
0.3286	10.0	15970	0.1937	0.9821	0.5000	0.5000	-1
0.326	11.0	17567	0.1813	0.9834	0.4993	0.4993	-1
0.3209	12.0	19164	0.1643	0.9821	0.4994	0.4994	-1
0.3152	13.0	20761	0.1795	0.9826	0.4984	0.4984	-1
0.3117	14.0	22358	0.1847	0.9826	0.4989	0.4989	-1
0.3057	15.0	23955	0.1881	0.9819	0.4999	0.4999	-1
0.2994	16.0	25552	0.1934	0.9829	0.4983	0.4983	-1
0.2969	17.0	27149	0.1774	0.9839	0.4985	0.4985	-1
0.2915	18.0	28746	0.1772	0.9836	0.4990	0.4990	-1
0.2877	19.0	30343	0.1781	0.9839	0.4992	0.4992	-1
0.2956	20.0	31940	0.1749	0.9839	0.4995	0.4995	-1

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.2+cu118
• Datasets 2.18.0
• Tokenizers 0.15.2