hubert-classifier-aug-ref

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: 3.1461
• Accuracy: 0.1671
• Precision: 0.0661
• Recall: 0.1671
• F1: 0.0830
• Binary: 0.4137

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: 1e-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
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	Binary
No log	0.19	50	4.4123	0.0377	0.0239	0.0377	0.0213	0.2075
No log	0.38	100	4.3574	0.0674	0.0177	0.0674	0.0253	0.2741
No log	0.58	150	4.2332	0.0323	0.0017	0.0323	0.0032	0.2884
No log	0.77	200	4.1388	0.0647	0.0160	0.0647	0.0182	0.3380
No log	0.96	250	4.0567	0.0674	0.0350	0.0674	0.0222	0.3407
No log	1.15	300	4.0043	0.0566	0.0114	0.0566	0.0143	0.3221
No log	1.34	350	3.9470	0.0485	0.0049	0.0485	0.0080	0.3221
No log	1.53	400	3.8803	0.0593	0.0124	0.0593	0.0135	0.3353
No log	1.73	450	3.8326	0.0566	0.0057	0.0566	0.0097	0.3323
4.1711	1.92	500	3.7760	0.0566	0.0061	0.0566	0.0103	0.3356
4.1711	2.11	550	3.7454	0.0647	0.0066	0.0647	0.0118	0.3372
4.1711	2.3	600	3.7036	0.0701	0.0075	0.0701	0.0132	0.3429
4.1711	2.49	650	3.6729	0.0728	0.0094	0.0728	0.0161	0.3431
4.1711	2.68	700	3.6306	0.0728	0.0117	0.0728	0.0177	0.3461
4.1711	2.88	750	3.6075	0.0836	0.0155	0.0836	0.0237	0.3536
4.1711	3.07	800	3.5817	0.0943	0.0284	0.0943	0.0285	0.3604
4.1711	3.26	850	3.5607	0.0916	0.0179	0.0916	0.0272	0.3577
4.1711	3.45	900	3.5373	0.0943	0.0214	0.0943	0.0304	0.3588
4.1711	3.64	950	3.5083	0.1078	0.0357	0.1078	0.0464	0.3714
3.7424	3.84	1000	3.4717	0.1105	0.0512	0.1105	0.0520	0.3765
3.7424	4.03	1050	3.4619	0.1213	0.0361	0.1213	0.0489	0.3825
3.7424	4.22	1100	3.4375	0.1240	0.0453	0.1240	0.0554	0.3844
3.7424	4.41	1150	3.4282	0.1267	0.0390	0.1267	0.0547	0.3849
3.7424	4.6	1200	3.4076	0.1267	0.0334	0.1267	0.0493	0.3838
3.7424	4.79	1250	3.3875	0.1078	0.0263	0.1078	0.0388	0.3730
3.7424	4.99	1300	3.3746	0.1240	0.0547	0.1240	0.0496	0.3822
3.7424	5.18	1350	3.3459	0.1375	0.0621	0.1375	0.0618	0.3946
3.7424	5.37	1400	3.3313	0.1375	0.0598	0.1375	0.0650	0.3946
3.7424	5.56	1450	3.3263	0.1429	0.0556	0.1429	0.0623	0.3951
3.5358	5.75	1500	3.3100	0.1348	0.0629	0.1348	0.0640	0.3895
3.5358	5.94	1550	3.2880	0.1402	0.0637	0.1402	0.0641	0.3957
3.5358	6.14	1600	3.2742	0.1402	0.0628	0.1402	0.0640	0.3965
3.5358	6.33	1650	3.2605	0.1509	0.0861	0.1509	0.0786	0.4049
3.5358	6.52	1700	3.2480	0.1429	0.0626	0.1429	0.0663	0.3976
3.5358	6.71	1750	3.2435	0.1482	0.0575	0.1482	0.0665	0.4030
3.5358	6.9	1800	3.2324	0.1482	0.0619	0.1482	0.0670	0.4022
3.5358	7.09	1850	3.2193	0.1563	0.0806	0.1563	0.0799	0.4070
3.5358	7.29	1900	3.2122	0.1644	0.0825	0.1644	0.0865	0.4119
3.5358	7.48	1950	3.1995	0.1617	0.0776	0.1617	0.0836	0.4108
3.4065	7.67	2000	3.1945	0.1617	0.0771	0.1617	0.0837	0.4116
3.4065	7.86	2050	3.1851	0.1725	0.0832	0.1725	0.0919	0.4191
3.4065	8.05	2100	3.1805	0.1617	0.0592	0.1617	0.0776	0.4100
3.4065	8.25	2150	3.1729	0.1617	0.0573	0.1617	0.0762	0.4100
3.4065	8.44	2200	3.1696	0.1617	0.0571	0.1617	0.0750	0.4100
3.4065	8.63	2250	3.1638	0.1644	0.0651	0.1644	0.0781	0.4119
3.4065	8.82	2300	3.1597	0.1590	0.0540	0.1590	0.0735	0.4089
3.4065	9.01	2350	3.1548	0.1671	0.0688	0.1671	0.0860	0.4137
3.4065	9.2	2400	3.1540	0.1617	0.0623	0.1617	0.0798	0.4100
3.4065	9.4	2450	3.1489	0.1644	0.0661	0.1644	0.0820	0.4119
3.3382	9.59	2500	3.1493	0.1644	0.0706	0.1644	0.0820	0.4119
3.3382	9.78	2550	3.1464	0.1671	0.0661	0.1671	0.0831	0.4137
3.3382	9.97	2600	3.1461	0.1671	0.0661	0.1671	0.0830	0.4137

Framework versions

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