dit-tiny_tobacco3482_kd_CEKD_t5.0_a0.7

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: 3.1844
• Accuracy: 0.18
• Brier Loss: 0.8763
• Nll: 6.0873
• F1 Micro: 0.18
• F1 Macro: 0.0306
• Ece: 0.2492
• Aurc: 0.8505

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	3.3625	0.145	0.8999	10.1577	0.145	0.0253	0.2220	0.8466
No log	1.96	6	3.3300	0.145	0.8947	10.5652	0.145	0.0253	0.2237	0.8468
No log	2.96	9	3.2822	0.14	0.8870	8.5877	0.14	0.0453	0.2040	0.8325
No log	3.96	12	3.2442	0.16	0.8812	6.5385	0.16	0.0327	0.2208	0.8814
No log	4.96	15	3.2219	0.155	0.8784	7.1527	0.155	0.0271	0.2352	0.8898
No log	5.96	18	3.2105	0.185	0.8778	8.7319	0.185	0.0517	0.2548	0.8944
No log	6.96	21	3.2032	0.18	0.8778	8.8034	0.18	0.0308	0.2478	0.8527
No log	7.96	24	3.1980	0.18	0.8779	8.1814	0.18	0.0306	0.2635	0.8527
No log	8.96	27	3.1937	0.18	0.8777	7.0314	0.18	0.0306	0.2618	0.8529
No log	9.96	30	3.1915	0.18	0.8776	6.9166	0.18	0.0306	0.2591	0.8537
No log	10.96	33	3.1900	0.18	0.8774	6.8864	0.18	0.0306	0.2551	0.8535
No log	11.96	36	3.1889	0.18	0.8773	6.5148	0.18	0.0306	0.2547	0.8532
No log	12.96	39	3.1881	0.18	0.8771	6.1469	0.18	0.0306	0.2543	0.8530
No log	13.96	42	3.1872	0.18	0.8769	6.1318	0.18	0.0306	0.2538	0.8525
No log	14.96	45	3.1865	0.18	0.8768	6.0783	0.18	0.0306	0.2501	0.8525
No log	15.96	48	3.1859	0.18	0.8766	6.0654	0.18	0.0306	0.2500	0.8520
No log	16.96	51	3.1855	0.18	0.8766	6.0809	0.18	0.0306	0.2459	0.8516
No log	17.96	54	3.1855	0.18	0.8766	6.0610	0.18	0.0306	0.2497	0.8515
No log	18.96	57	3.1854	0.18	0.8766	6.0659	0.18	0.0306	0.2579	0.8515
No log	19.96	60	3.1850	0.18	0.8764	6.0737	0.18	0.0306	0.2656	0.8513
No log	20.96	63	3.1848	0.18	0.8764	6.0869	0.18	0.0306	0.2575	0.8510
No log	21.96	66	3.1846	0.18	0.8764	6.1423	0.18	0.0306	0.2533	0.8510
No log	22.96	69	3.1845	0.18	0.8763	6.1047	0.18	0.0306	0.2532	0.8505
No log	23.96	72	3.1845	0.18	0.8763	6.0895	0.18	0.0306	0.2532	0.8504
No log	24.96	75	3.1844	0.18	0.8763	6.0873	0.18	0.0306	0.2492	0.8505

Framework versions

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