jialicheng/unlearn_speech_commands_hubert-large_neggrad_2_42

superb_ks_42

This model is a fine-tuned version of facebook/hubert-large-ll60k on the superb dataset. It achieves the following results on the evaluation set:

• Loss: -6.8759
• Accuracy: 0.6209
• Test Accuracy: 0.6209
• Df Accuracy: 0.1352
• Unlearn Overall Accuracy: 0.7429
• Unlearn Time: 652.0674

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: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Overall Accuracy	Unlearn Overall Accuracy	Time
No log	1.0	32	-0.1223	0.9598	0.5101	0.5101	-1
No log	2.0	64	-6.8759	0.1352	0.7429	0.7429	-1
No log	3.0	96	-9.1867	0.1352	0.7429	0.7429	-1
No log	4.0	128	-11.1371	0.1352	0.7429	0.7429	-1
No log	5.0	160	-12.8684	0.1352	0.7429	0.7429	-1
No log	6.0	192	-14.3520	0.1352	0.7429	0.7429	-1
No log	7.0	224	-15.5595	0.1352	0.7429	0.7429	-1
No log	8.0	256	-16.4513	0.1352	0.7429	0.7429	-1
No log	9.0	288	-16.9985	0.1352	0.7429	0.7429	-1
No log	10.0	320	-17.1870	0.1352	0.7429	0.7429	-1

Framework versions

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