galsenai/hubert-large-ls960-ft-waxal-keyword-spotting

hubert-large-ls960-ft

This model is a fine-tuned version of facebook/hubert-large-ls960-ft on the galsenai/waxal_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.3272
• Accuracy: 0.9413
• Precision: 0.9865
• F1: 0.9628

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: 12
• eval_batch_size: 12
• seed: 0
• gradient_accumulation_steps: 4
• total_train_batch_size: 48
• 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
4.7142	1.01	500	5.2765	0.0	0.0	0.0
4.396	2.02	1000	5.4145	0.0	0.0	0.0
3.8883	3.04	1500	4.4336	0.0474	0.0408	0.0104
2.7848	4.05	2000	3.9772	0.1300	0.1281	0.0964
1.8649	5.06	2500	3.4482	0.1576	0.3339	0.1547
1.3084	6.07	3000	2.9703	0.3081	0.5296	0.3402
0.9868	7.08	3500	2.3985	0.4687	0.8032	0.5353
0.7679	8.1	4000	1.7937	0.6521	0.8389	0.7095
0.6232	9.11	4500	1.4768	0.7389	0.8698	0.7847
0.5126	10.12	5000	1.0542	0.8287	0.9443	0.8763
0.4453	11.13	5500	0.9050	0.8518	0.9511	0.8960
0.3775	12.15	6000	0.6996	0.8928	0.9662	0.9266
0.3568	13.16	6500	0.6157	0.8958	0.9743	0.9285
0.3165	14.17	7000	0.4925	0.9151	0.9764	0.9436
0.2951	15.18	7500	0.4992	0.9038	0.9773	0.9369
0.2763	16.19	8000	0.5212	0.9072	0.9821	0.9404
0.2634	17.21	8500	0.5201	0.9087	0.9817	0.9418
0.2422	18.22	9000	0.4504	0.9235	0.9840	0.9514
0.236	19.23	9500	0.3829	0.9257	0.9825	0.9518
0.2272	20.24	10000	0.4632	0.9155	0.9822	0.9451
0.226	21.25	10500	0.4731	0.9159	0.9837	0.9470
0.2129	22.27	11000	0.3814	0.9299	0.9832	0.9549
0.2009	23.28	11500	0.4119	0.9257	0.9814	0.9515
0.1973	24.29	12000	0.4310	0.9216	0.9843	0.9493
0.1965	25.3	12500	0.3272	0.9413	0.9865	0.9628
0.1989	26.32	13000	0.4231	0.9242	0.9878	0.9528
0.1916	27.33	13500	0.3978	0.9284	0.9876	0.9559
0.1849	28.34	14000	0.4529	0.9216	0.9865	0.9507
0.1844	29.35	14500	0.3854	0.9314	0.9864	0.9566
0.1831	30.36	15000	0.4178	0.9257	0.9853	0.9528
0.1778	31.38	15500	0.3737	0.9360	0.9884	0.9606

Framework versions

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