wav2vec2-base-sound2

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

• Loss: 1.5012
• Accuracy: 0.5357

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: 9e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	1	2.0762	0.0714
No log	2.0	2	2.0638	0.1429
No log	3.0	3	2.0387	0.2143
No log	4.0	4	2.0124	0.2143
No log	5.0	5	1.9864	0.2143
No log	6.0	6	1.9609	0.2143
No log	7.0	7	1.9235	0.2143
No log	8.0	8	1.9379	0.2143
No log	9.0	9	1.8627	0.2857
1.9713	10.0	10	1.8277	0.3214
1.9713	11.0	11	1.7765	0.3571
1.9713	12.0	12	1.7204	0.5
1.9713	13.0	13	1.6956	0.5
1.9713	14.0	14	1.6602	0.5357
1.9713	15.0	15	1.6277	0.5714
1.9713	16.0	16	1.6053	0.5
1.9713	17.0	17	1.5825	0.5
1.9713	18.0	18	1.5656	0.4286
1.9713	19.0	19	1.5616	0.4643
1.6334	20.0	20	1.5613	0.4286
1.6334	21.0	21	1.5419	0.5
1.6334	22.0	22	1.5166	0.5357
1.6334	23.0	23	1.5088	0.5
1.6334	24.0	24	1.5052	0.5
1.6334	25.0	25	1.5012	0.5357

Framework versions

• Transformers 4.18.0
• Pytorch 1.11.0+cu113
• Datasets 1.14.0
• Tokenizers 0.12.1