dit-small_tobacco3482_kd_MSE

This model is a fine-tuned version of microsoft/dit-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 6.7275
• Accuracy: 0.21
• Brier Loss: 0.8834
• Nll: 6.7677
• F1 Micro: 0.2100
• F1 Macro: 0.1146
• Ece: 0.2647
• Aurc: 0.7666

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Brier Loss	Nll	F1 Micro	F1 Macro	Ece	Aurc
No log	0.96	3	7.1014	0.06	0.9055	7.9056	0.06	0.0114	0.1732	0.9050
No log	1.96	6	6.9659	0.125	0.8970	10.1253	0.125	0.0631	0.2010	0.8465
No log	2.96	9	6.8528	0.075	0.8954	7.0315	0.075	0.0258	0.1912	0.8871
No log	3.96	12	6.8522	0.205	0.8955	7.0990	0.205	0.0776	0.2426	0.7588
No log	4.96	15	6.8465	0.19	0.8959	7.1340	0.19	0.0627	0.2308	0.7536
No log	5.96	18	6.8246	0.205	0.8937	7.1101	0.205	0.0867	0.2410	0.7354
No log	6.96	21	6.8054	0.085	0.8918	7.0215	0.085	0.0435	0.1847	0.8289
No log	7.96	24	6.8025	0.22	0.8879	6.8272	0.22	0.0967	0.2487	0.7438
No log	8.96	27	6.8045	0.21	0.8871	6.3740	0.2100	0.0992	0.2412	0.7634
No log	9.96	30	6.8013	0.22	0.8869	6.9538	0.22	0.1016	0.2495	0.7633
No log	10.96	33	6.7920	0.215	0.8865	6.9670	0.2150	0.0968	0.2549	0.7577
No log	11.96	36	6.7817	0.22	0.8867	6.9953	0.22	0.1004	0.2455	0.7437
No log	12.96	39	6.7729	0.17	0.8884	6.9738	0.17	0.0891	0.2277	0.7865
No log	13.96	42	6.7632	0.2	0.8873	6.9622	0.2000	0.0998	0.2393	0.7413
No log	14.96	45	6.7548	0.215	0.8860	6.9576	0.2150	0.1010	0.2635	0.7189
No log	15.96	48	6.7489	0.22	0.8857	6.8386	0.22	0.1024	0.2665	0.7098
No log	16.96	51	6.7457	0.23	0.8855	6.8730	0.23	0.1129	0.2506	0.7217
No log	17.96	54	6.7455	0.215	0.8864	6.8688	0.2150	0.1058	0.2576	0.7528
No log	18.96	57	6.7424	0.16	0.8861	6.8631	0.16	0.0843	0.2281	0.8036
No log	19.96	60	6.7380	0.155	0.8850	6.8443	0.155	0.0871	0.2315	0.7937
No log	20.96	63	6.7348	0.195	0.8841	6.7769	0.195	0.0949	0.2501	0.7799
No log	21.96	66	6.7317	0.175	0.8838	6.7692	0.175	0.1025	0.2421	0.7797
No log	22.96	69	6.7293	0.175	0.8836	6.7682	0.175	0.1012	0.2452	0.7799
No log	23.96	72	6.7281	0.205	0.8834	6.7672	0.205	0.1132	0.2566	0.7679
No log	24.96	75	6.7275	0.21	0.8834	6.7677	0.2100	0.1146	0.2647	0.7666

Framework versions

• Transformers 4.26.1
• Pytorch 1.13.1.post200
• Datasets 2.9.0
• Tokenizers 0.13.2