hubert-classifier-aug

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: 0.5817
• Accuracy: 0.8356
• Precision: 0.8647
• Recall: 0.8356
• F1: 0.8286
• Binary: 0.8852

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: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	Binary
No log	0.19	50	4.0265	0.0404	0.0071	0.0404	0.0079	0.3011
No log	0.38	100	3.5428	0.0485	0.0048	0.0485	0.0080	0.3286
No log	0.58	150	3.3405	0.0836	0.0175	0.0836	0.0251	0.3555
No log	0.77	200	3.2238	0.0809	0.0143	0.0809	0.0216	0.3512
No log	0.96	250	3.1041	0.0728	0.0135	0.0728	0.0202	0.3493
No log	1.15	300	2.9851	0.1078	0.0517	0.1078	0.0480	0.3730
No log	1.34	350	2.8525	0.1779	0.0780	0.1779	0.0876	0.4197
No log	1.53	400	2.7647	0.1752	0.1108	0.1752	0.1064	0.4221
No log	1.73	450	2.5521	0.2291	0.1539	0.2291	0.1450	0.4547
3.3693	1.92	500	2.4121	0.2372	0.1655	0.2372	0.1618	0.4668
3.3693	2.11	550	2.2312	0.2992	0.2286	0.2992	0.2185	0.5081
3.3693	2.3	600	2.0065	0.4124	0.2985	0.4124	0.3133	0.5865
3.3693	2.49	650	1.8816	0.4313	0.3359	0.4313	0.3461	0.6013
3.3693	2.68	700	1.8069	0.4906	0.4702	0.4906	0.4308	0.6426
3.3693	2.88	750	1.6310	0.5418	0.4981	0.5418	0.4728	0.6803
3.3693	3.07	800	1.5274	0.5580	0.5219	0.5580	0.5002	0.6908
3.3693	3.26	850	1.3417	0.6415	0.6343	0.6415	0.5980	0.7544
3.3693	3.45	900	1.3121	0.6173	0.6059	0.6173	0.5690	0.7334
3.3693	3.64	950	1.2298	0.6523	0.6501	0.6523	0.6183	0.7577
2.2303	3.84	1000	1.1427	0.7197	0.7323	0.7197	0.6897	0.8040
2.2303	4.03	1050	1.0947	0.6765	0.6891	0.6765	0.6387	0.7741
2.2303	4.22	1100	1.1233	0.6361	0.6473	0.6361	0.6054	0.7447
2.2303	4.41	1150	0.9765	0.7547	0.7606	0.7547	0.7331	0.8296
2.2303	4.6	1200	0.9206	0.7547	0.7546	0.7547	0.7270	0.8305
2.2303	4.79	1250	0.8658	0.7790	0.7868	0.7790	0.7625	0.8456
2.2303	4.99	1300	0.8961	0.7385	0.7576	0.7385	0.7254	0.8186
2.2303	5.18	1350	0.7709	0.8005	0.8185	0.8005	0.7912	0.8596
2.2303	5.37	1400	0.7638	0.7925	0.8118	0.7925	0.7760	0.8547
2.2303	5.56	1450	0.7085	0.8194	0.8415	0.8194	0.8081	0.8741
1.6078	5.75	1500	0.7230	0.7790	0.8195	0.7790	0.7739	0.8456
1.6078	5.94	1550	0.6475	0.7951	0.8174	0.7951	0.7813	0.8558
1.6078	6.14	1600	0.6910	0.7844	0.8082	0.7844	0.7686	0.8504
1.6078	6.33	1650	0.6233	0.8194	0.8462	0.8194	0.8111	0.8730
1.6078	6.52	1700	0.6599	0.8059	0.8429	0.8059	0.8031	0.8633
1.6078	6.71	1750	0.6999	0.7925	0.8119	0.7925	0.7751	0.8550
1.6078	6.9	1800	0.6271	0.8140	0.8266	0.8140	0.8018	0.8701
1.6078	7.09	1850	0.5545	0.8329	0.8557	0.8329	0.8288	0.8822
1.6078	7.29	1900	0.6343	0.8032	0.8179	0.8032	0.7930	0.8625
1.6078	7.48	1950	0.6007	0.8194	0.8447	0.8194	0.8136	0.8728
1.2974	7.67	2000	0.5878	0.8356	0.8674	0.8356	0.8333	0.8841
1.2974	7.86	2050	0.6410	0.8086	0.8344	0.8086	0.8011	0.8652
1.2974	8.05	2100	0.6430	0.8005	0.8201	0.8005	0.7894	0.8598
1.2974	8.25	2150	0.5540	0.8221	0.8414	0.8221	0.8177	0.8747
1.2974	8.44	2200	0.5511	0.8356	0.8635	0.8356	0.8317	0.8833
1.2974	8.63	2250	0.5817	0.8356	0.8647	0.8356	0.8286	0.8852

Framework versions

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