jialicheng/unlearn_speech_commands_whisper-base_neggrad_6_42

superb_ks_42

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

• Loss: -9.7860
• Accuracy: 0.6209
• Test Accuracy: 0.6209
• Df Accuracy: 0.1331
• Unlearn Overall Accuracy: 0.7439
• Unlearn Time: 1310.8787

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	96	-9.7860	0.1331	0.7439	0.7439	-1
No log	2.0	192	-29.0068	0.1331	0.7439	0.7439	-1
No log	3.0	288	-51.2519	0.1331	0.7439	0.7439	-1
No log	4.0	384	-74.5230	0.1331	0.7439	0.7439	-1
No log	5.0	480	-96.9840	0.1331	0.7439	0.7439	-1
No log	6.0	576	-117.2810	0.1331	0.7439	0.7439	-1
No log	7.0	672	-134.2495	0.1331	0.7439	0.7439	-1
No log	8.0	768	-147.0133	0.1331	0.7439	0.7439	-1
No log	9.0	864	-154.9318	0.1331	0.7439	0.7439	-1
No log	10.0	960	-157.6337	0.1331	0.7439	0.7439	-1

Framework versions

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