jialicheng/unlearn_speech_commands_hubert-base_random_label_6_42

superb_ks_42

This model is a fine-tuned version of facebook/hubert-base-ls960 on the superb dataset. It achieves the following results on the evaluation set:

• Loss: 0.1802
• Accuracy: 0.9728
• Test Accuracy: 0.9728
• Df Accuracy: 0.9445
• Unlearn Overall Accuracy: 0.5141
• Unlearn Time: 3235.3319

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
0.4113	1.0	1597	0.1802	0.9445	0.5141	0.5141	-1
0.2928	2.0	3194	0.1519	0.9775	0.5015	0.5015	-1
0.2926	3.0	4791	0.1724	0.9690	0.5038	0.5038	-1
0.2928	4.0	6388	0.1532	0.9788	0.5021	0.5021	-1
0.2777	5.0	7985	0.1771	0.9778	0.5009	0.5009	-1
0.2839	6.0	9582	0.1589	0.9785	0.5019	0.5019	-1
0.2706	7.0	11179	0.1333	0.9798	0.5022	0.5022	-1
0.2668	8.0	12776	0.1606	0.9791	0.5007	0.5007	-1
0.2602	9.0	14373	0.1639	0.9775	0.5018	0.5018	-1
0.2571	10.0	15970	0.1563	0.9772	0.5031	0.5031	-1
0.2555	11.0	17567	0.1620	0.9808	0.5007	0.5007	-1
0.2501	12.0	19164	0.1482	0.9808	0.5009	0.5009	-1
0.245	13.0	20761	0.1625	0.9817	0.5005	0.5005	-1
0.2452	14.0	22358	0.1709	0.9791	0.5019	0.5019	-1
0.241	15.0	23955	0.1555	0.9804	0.5011	0.5011	-1
0.2375	16.0	25552	0.1414	0.9801	0.5024	0.5024	-1
0.2303	17.0	27149	0.1627	0.9788	0.5022	0.5022	-1
0.2274	18.0	28746	0.1526	0.9788	0.5024	0.5024	-1
0.2263	19.0	30343	0.1547	0.9801	0.5015	0.5015	-1
0.2305	20.0	31940	0.1557	0.9798	0.5020	0.5020	-1

Framework versions

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