hubert-classifier-aug-fold-1

This model is a fine-tuned version of facebook/hubert-base-ls960 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5322
• Accuracy: 0.8693
• Precision: 0.8831
• Recall: 0.8693
• F1: 0.8678
• Binary: 0.9088

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: 0.0001
• 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
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	Binary
No log	0.13	50	3.9624	0.0351	0.0015	0.0351	0.0028	0.3143
No log	0.27	100	3.4385	0.0634	0.0224	0.0634	0.0195	0.3432
No log	0.4	150	3.2477	0.0864	0.0446	0.0864	0.0302	0.3568
No log	0.54	200	2.9937	0.1282	0.0531	0.1282	0.0622	0.3873
No log	0.67	250	2.7511	0.2456	0.1513	0.2456	0.1564	0.4709
No log	0.81	300	2.4666	0.3077	0.2328	0.3077	0.2207	0.5151
No log	0.94	350	2.1657	0.3954	0.3244	0.3954	0.3091	0.5765
3.2407	1.08	400	1.8980	0.4453	0.3778	0.4453	0.3735	0.6128
3.2407	1.21	450	1.6861	0.5371	0.4729	0.5371	0.4709	0.6744
3.2407	1.35	500	1.5060	0.5897	0.5801	0.5897	0.5439	0.7140
3.2407	1.48	550	1.3841	0.6019	0.5868	0.6019	0.5615	0.7189
3.2407	1.62	600	1.2200	0.6640	0.6817	0.6640	0.6403	0.7637
3.2407	1.75	650	1.1366	0.6869	0.7098	0.6869	0.6628	0.7799
3.2407	1.89	700	0.9917	0.7382	0.7498	0.7382	0.7264	0.8174
1.6581	2.02	750	0.9604	0.7355	0.7688	0.7355	0.7248	0.8144
1.6581	2.16	800	0.8836	0.7679	0.7816	0.7679	0.7596	0.8391
1.6581	2.29	850	0.8232	0.7787	0.8122	0.7787	0.7725	0.8463
1.6581	2.43	900	0.7861	0.7800	0.8017	0.7800	0.7711	0.8467
1.6581	2.56	950	0.7454	0.8030	0.8288	0.8030	0.7955	0.8632
1.6581	2.7	1000	0.6824	0.8178	0.8450	0.8178	0.8135	0.8726
1.6581	2.83	1050	0.7187	0.8138	0.8429	0.8138	0.8116	0.8714
1.6581	2.96	1100	0.7004	0.8084	0.8280	0.8084	0.8041	0.8675
0.9905	3.1	1150	0.6490	0.8219	0.8390	0.8219	0.8176	0.8771
0.9905	3.23	1200	0.6872	0.8178	0.8328	0.8178	0.8154	0.8737
0.9905	3.37	1250	0.6676	0.8340	0.8548	0.8340	0.8270	0.8850
0.9905	3.5	1300	0.6439	0.8205	0.8480	0.8205	0.8212	0.8750
0.9905	3.64	1350	0.5648	0.8354	0.8527	0.8354	0.8319	0.8854
0.9905	3.77	1400	0.6231	0.8340	0.8537	0.8340	0.8310	0.8854
0.9905	3.91	1450	0.6813	0.8232	0.8532	0.8232	0.8219	0.8779
0.7084	4.04	1500	0.6047	0.8475	0.8680	0.8475	0.8459	0.8941
0.7084	4.18	1550	0.6319	0.8354	0.8489	0.8354	0.8323	0.8869
0.7084	4.31	1600	0.5892	0.8650	0.8814	0.8650	0.8625	0.9066
0.7084	4.45	1650	0.5572	0.8650	0.8825	0.8650	0.8649	0.9066
0.7084	4.58	1700	0.6023	0.8556	0.8705	0.8556	0.8530	0.9005
0.7084	4.72	1750	0.5772	0.8489	0.8729	0.8489	0.8456	0.8953
0.7084	4.85	1800	0.6242	0.8435	0.8643	0.8435	0.8416	0.8920
0.7084	4.99	1850	0.5903	0.8421	0.8574	0.8421	0.8392	0.8906
0.5634	5.12	1900	0.6256	0.8475	0.8598	0.8475	0.8438	0.8949
0.5634	5.26	1950	0.5891	0.8556	0.8728	0.8556	0.8553	0.9005
0.5634	5.39	2000	0.5435	0.8704	0.8843	0.8704	0.8681	0.9100
0.5634	5.53	2050	0.5093	0.8853	0.8952	0.8853	0.8840	0.9204
0.5634	5.66	2100	0.6058	0.8677	0.8784	0.8677	0.8676	0.9086
0.5634	5.8	2150	0.5635	0.8610	0.8728	0.8610	0.8607	0.9040
0.5634	5.93	2200	0.5897	0.8664	0.8828	0.8664	0.8661	0.9077
0.466	6.06	2250	0.6280	0.8623	0.8830	0.8623	0.8612	0.9058
0.466	6.2	2300	0.7129	0.8394	0.8548	0.8394	0.8371	0.8888
0.466	6.33	2350	0.6993	0.8435	0.8606	0.8435	0.8408	0.8926
0.466	6.47	2400	0.6314	0.8462	0.8630	0.8462	0.8440	0.8934

Framework versions

• Transformers 4.38.2
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.15.1