galsenai/wavlm-base-waxal-keyword-spotting

wavlm-base

This model is a fine-tuned version of microsoft/wavlm-base on the galsenai/waxal_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 2.1345
• Accuracy: 0.6783
• Precision: 0.8774
• F1: 0.7615

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: 3e-05
• train_batch_size: 30
• eval_batch_size: 30
• seed: 0
• gradient_accumulation_steps: 4
• total_train_batch_size: 120
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 32.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	F1
4.4506	2.53	500	4.8601	0.0224	0.0136	0.0066
3.0523	5.05	1000	4.6674	0.0720	0.0460	0.0394
1.949	7.58	1500	4.1533	0.1156	0.1847	0.1064
1.3427	10.1	2000	3.8173	0.1448	0.2382	0.1347
1.0064	12.63	2500	3.5546	0.2183	0.4464	0.2385
0.7985	15.15	3000	3.1172	0.3842	0.6336	0.4258
0.6505	17.68	3500	2.9231	0.5165	0.7677	0.5995
0.5367	20.2	4000	2.4935	0.5961	0.8182	0.6755
0.465	22.73	4500	2.2411	0.6412	0.8624	0.7272
0.4075	25.25	5000	2.1345	0.6783	0.8774	0.7615
0.3793	27.78	5500	2.2535	0.6681	0.8792	0.7543
0.3418	30.3	6000	2.3390	0.6662	0.8905	0.7576

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.11.0+cu113
• Datasets 2.9.1.dev0
• Tokenizers 0.13.2