dit-tiny_tobacco3482_kd_CEKD_t2.0_a0.5

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: 3.5976
• Accuracy: 0.18
• Brier Loss: 0.8781
• Nll: 6.8947
• F1 Micro: 0.18
• F1 Macro: 0.0306
• Ece: 0.2499
• Aurc: 0.8510

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	3.8479	0.145	0.8999	10.1604	0.145	0.0253	0.2222	0.8467
No log	1.96	6	3.8090	0.145	0.8946	10.5967	0.145	0.0253	0.2246	0.8470
No log	2.96	9	3.7500	0.16	0.8866	8.6365	0.16	0.0406	0.2205	0.8486
No log	3.96	12	3.7003	0.16	0.8805	6.5484	0.16	0.0327	0.2242	0.8816
No log	4.96	15	3.6677	0.155	0.8776	6.7592	0.155	0.0271	0.2365	0.8919
No log	5.96	18	3.6477	0.155	0.8770	7.2639	0.155	0.0278	0.2368	0.8961
No log	6.96	21	3.6339	0.18	0.8774	7.3546	0.18	0.0313	0.2486	0.8556
No log	7.96	24	3.6240	0.18	0.8781	7.0685	0.18	0.0308	0.2654	0.8528
No log	8.96	27	3.6163	0.18	0.8784	7.0041	0.18	0.0306	0.2561	0.8532
No log	9.96	30	3.6114	0.18	0.8787	6.9904	0.18	0.0306	0.2584	0.8537
No log	10.96	33	3.6078	0.18	0.8788	6.9806	0.18	0.0306	0.2594	0.8538
No log	11.96	36	3.6052	0.18	0.8789	6.9768	0.18	0.0306	0.2596	0.8537
No log	12.96	39	3.6034	0.18	0.8788	6.9716	0.18	0.0306	0.2507	0.8532
No log	13.96	42	3.6018	0.18	0.8786	6.9683	0.18	0.0306	0.2548	0.8527
No log	14.96	45	3.6005	0.18	0.8786	6.9040	0.18	0.0306	0.2597	0.8524
No log	15.96	48	3.5995	0.18	0.8784	6.8978	0.18	0.0306	0.2685	0.8518
No log	16.96	51	3.5989	0.18	0.8784	6.8972	0.18	0.0306	0.2641	0.8515
No log	17.96	54	3.5989	0.18	0.8784	6.8961	0.18	0.0306	0.2550	0.8513
No log	18.96	57	3.5988	0.18	0.8784	6.8968	0.18	0.0306	0.2505	0.8510
No log	19.96	60	3.5982	0.18	0.8782	6.8956	0.18	0.0306	0.2478	0.8511
No log	20.96	63	3.5980	0.18	0.8782	6.8954	0.18	0.0306	0.2456	0.8507
No log	21.96	66	3.5978	0.18	0.8782	6.8951	0.18	0.0306	0.2499	0.8511
No log	22.96	69	3.5976	0.18	0.8781	6.8949	0.18	0.0306	0.2499	0.8510
No log	23.96	72	3.5976	0.18	0.8781	6.8949	0.18	0.0306	0.2499	0.8510
No log	24.96	75	3.5976	0.18	0.8781	6.8947	0.18	0.0306	0.2499	0.8510

Framework versions

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