dit-tiny_tobacco3482_kd_CEKD_t1.5_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: 2.6280
• Accuracy: 0.18
• Brier Loss: 0.8747
• Nll: 6.7569
• F1 Micro: 0.18
• F1 Macro: 0.0306
• Ece: 0.2550
• Aurc: 0.8496

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	2.7961	0.145	0.8999	10.1560	0.145	0.0253	0.2221	0.8467
No log	1.96	6	2.7646	0.145	0.8946	10.5828	0.145	0.0253	0.2242	0.8475
No log	2.96	9	2.7185	0.155	0.8868	8.6137	0.155	0.0501	0.2145	0.8394
No log	3.96	12	2.6825	0.21	0.8808	6.5439	0.2100	0.0613	0.2567	0.8351
No log	4.96	15	2.6619	0.155	0.8778	6.7839	0.155	0.0274	0.2346	0.8880
No log	5.96	18	2.6517	0.18	0.8769	7.4578	0.18	0.0395	0.2461	0.8571
No log	6.96	21	2.6450	0.18	0.8767	7.1192	0.18	0.0308	0.2518	0.8516
No log	7.96	24	2.6400	0.18	0.8766	6.9539	0.18	0.0306	0.2472	0.8526
No log	8.96	27	2.6355	0.18	0.8762	6.9109	0.18	0.0306	0.2524	0.8527
No log	9.96	30	2.6332	0.18	0.8759	6.8997	0.18	0.0306	0.2491	0.8527
No log	10.96	33	2.6317	0.18	0.8757	6.8943	0.18	0.0306	0.2529	0.8524
No log	11.96	36	2.6309	0.18	0.8755	6.8287	0.18	0.0306	0.2442	0.8523
No log	12.96	39	2.6304	0.18	0.8753	6.7670	0.18	0.0306	0.2478	0.8521
No log	13.96	42	2.6298	0.18	0.8752	6.7597	0.18	0.0306	0.2433	0.8517
No log	14.96	45	2.6293	0.18	0.8751	6.7590	0.18	0.0306	0.2516	0.8513
No log	15.96	48	2.6290	0.18	0.8750	6.7556	0.18	0.0306	0.2555	0.8515
No log	16.96	51	2.6287	0.18	0.8750	6.7582	0.18	0.0306	0.2557	0.8514
No log	17.96	54	2.6289	0.18	0.8750	6.7556	0.18	0.0306	0.2476	0.8509
No log	18.96	57	2.6289	0.18	0.8750	6.7567	0.18	0.0306	0.2475	0.8505
No log	19.96	60	2.6285	0.18	0.8748	6.7567	0.18	0.0306	0.2433	0.8502
No log	20.96	63	2.6283	0.18	0.8748	6.7577	0.18	0.0306	0.2512	0.8500
No log	21.96	66	2.6281	0.18	0.8748	6.7586	0.18	0.0306	0.2551	0.8495
No log	22.96	69	2.6280	0.18	0.8747	6.7580	0.18	0.0306	0.2550	0.8496
No log	23.96	72	2.6280	0.18	0.8747	6.7573	0.18	0.0306	0.2550	0.8496
No log	24.96	75	2.6280	0.18	0.8747	6.7569	0.18	0.0306	0.2550	0.8496

Framework versions

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