jialicheng/unlearn_speech_commands_hubert-base_random_label_4_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: 0.1481
• Accuracy: 0.9790
• Test Accuracy: 0.9790
• Df Accuracy: 0.9736
• Unlearn Overall Accuracy: 0.5027
• Unlearn Time: 3242.5693

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
0.3018	1.0	1597	0.1200	0.9765	0.5025	0.5025	-1
0.2447	2.0	3194	0.1311	0.9790	0.4995	0.4995	-1
0.25	3.0	4791	0.1481	0.9736	0.5027	0.5027	-1
0.2432	4.0	6388	0.1275	0.9785	0.5014	0.5014	-1
0.238	5.0	7985	0.1632	0.9780	0.5020	0.5020	-1
0.229	6.0	9582	0.1489	0.9794	0.5013	0.5013	-1
0.2273	7.0	11179	0.1227	0.9794	0.5012	0.5012	-1
0.2229	8.0	12776	0.1194	0.9790	0.5016	0.5016	-1
0.2164	9.0	14373	0.1346	0.9799	0.5014	0.5014	-1
0.2112	10.0	15970	0.1497	0.9804	0.5007	0.5007	-1
0.2098	11.0	17567	0.1452	0.9809	0.5006	0.5006	-1
0.21	12.0	19164	0.1170	0.9804	0.5018	0.5018	-1
0.1941	13.0	20761	0.1271	0.9824	0.5006	0.5006	-1
0.1978	14.0	22358	0.1275	0.9804	0.5016	0.5016	-1
0.1975	15.0	23955	0.1395	0.9804	0.5006	0.5006	-1
0.1946	16.0	25552	0.1307	0.9809	0.5008	0.5008	-1
0.1862	17.0	27149	0.1301	0.9809	0.5009	0.5009	-1
0.1836	18.0	28746	0.1363	0.9809	0.5010	0.5010	-1
0.1754	19.0	30343	0.1348	0.9814	0.5005	0.5005	-1
0.1815	20.0	31940	0.1353	0.9809	0.5008	0.5008	-1

Framework versions

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