dit-small_tobacco3482_simkd_CEKD_t1_aNone

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: 0.9876
• Accuracy: 0.085
• Brier Loss: 0.8927
• Nll: 8.3272
• F1 Micro: 0.085
• F1 Macro: 0.0461
• Ece: 0.1645
• Aurc: 0.7988

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 64
• 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	12	1.0049	0.08	0.8993	5.4663	0.08	0.0322	0.1476	0.8883
No log	1.96	24	1.0007	0.165	0.8988	5.5926	0.165	0.0284	0.2066	0.8251
No log	2.96	36	0.9994	0.16	0.8982	5.9135	0.16	0.0277	0.2100	0.8518
No log	3.96	48	0.9984	0.17	0.8975	6.1195	0.17	0.0574	0.2142	0.8153
No log	4.96	60	0.9976	0.19	0.8970	6.2724	0.19	0.0752	0.2294	0.8254
No log	5.96	72	0.9967	0.09	0.8968	6.3787	0.09	0.0315	0.1591	0.7950
No log	6.96	84	0.9958	0.065	0.8964	6.4218	0.065	0.0122	0.1433	0.8333
No log	7.96	96	0.9949	0.065	0.8960	6.5170	0.065	0.0122	0.1543	0.8344
No log	8.96	108	0.9941	0.065	0.8956	6.5572	0.065	0.0123	0.1545	0.8331
No log	9.96	120	0.9934	0.07	0.8954	6.6362	0.07	0.0304	0.1597	0.8313
No log	10.96	132	0.9926	0.07	0.8951	6.6430	0.07	0.0304	0.1576	0.8325
No log	11.96	144	0.9920	0.07	0.8948	6.6842	0.07	0.0304	0.1590	0.8225
No log	12.96	156	0.9914	0.07	0.8947	6.7731	0.07	0.0304	0.1619	0.8155
No log	13.96	168	0.9909	0.07	0.8944	6.8584	0.07	0.0304	0.1522	0.8128
No log	14.96	180	0.9904	0.07	0.8941	6.8161	0.07	0.0304	0.1524	0.8142
No log	15.96	192	0.9899	0.07	0.8940	7.3169	0.07	0.0304	0.1532	0.8109
No log	16.96	204	0.9894	0.07	0.8937	7.8481	0.07	0.0304	0.1531	0.8132
No log	17.96	216	0.9890	0.08	0.8935	8.3375	0.08	0.0439	0.1587	0.8002
No log	18.96	228	0.9886	0.07	0.8933	8.4250	0.07	0.0307	0.1536	0.8132
No log	19.96	240	0.9883	0.085	0.8931	8.4316	0.085	0.0445	0.1618	0.8014
No log	20.96	252	0.9880	0.075	0.8930	8.4395	0.075	0.0392	0.1566	0.8088
No log	21.96	264	0.9878	0.085	0.8929	8.3319	0.085	0.0476	0.1621	0.7956
No log	22.96	276	0.9877	0.08	0.8928	8.3274	0.08	0.0439	0.1594	0.8024
No log	23.96	288	0.9876	0.08	0.8927	8.3285	0.08	0.0440	0.1595	0.8014
No log	24.96	300	0.9876	0.085	0.8927	8.3272	0.085	0.0461	0.1645	0.7988

Framework versions

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