dit-tiny_tobacco3482_kd_CEKD_t2.5_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: 3.9560
• Accuracy: 0.18
• Brier Loss: 0.8800
• Nll: 6.8606
• F1 Micro: 0.18
• F1 Macro: 0.0306
• Ece: 0.2612
• Aurc: 0.8512

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	4.2281	0.145	0.8999	10.1620	0.145	0.0253	0.2222	0.8467
No log	1.96	6	4.1872	0.145	0.8946	10.5915	0.145	0.0253	0.2275	0.8468
No log	2.96	9	4.1248	0.155	0.8866	8.6280	0.155	0.0360	0.2179	0.8487
No log	3.96	12	4.0716	0.155	0.8806	6.5480	0.155	0.0272	0.2254	0.8851
No log	4.96	15	4.0359	0.155	0.8778	6.7781	0.155	0.0271	0.2310	0.8931
No log	5.96	18	4.0135	0.155	0.8774	7.8547	0.155	0.0271	0.2345	0.8965
No log	6.96	21	3.9978	0.185	0.8779	8.3528	0.185	0.0468	0.2615	0.8612
No log	7.96	24	3.9867	0.18	0.8789	7.6001	0.18	0.0308	0.2618	0.8546
No log	8.96	27	3.9782	0.18	0.8796	7.0871	0.18	0.0306	0.2613	0.8538
No log	9.96	30	3.9726	0.18	0.8800	7.0519	0.18	0.0306	0.2687	0.8545
No log	10.96	33	3.9684	0.18	0.8803	7.0277	0.18	0.0306	0.2656	0.8537
No log	11.96	36	3.9654	0.18	0.8805	7.0162	0.18	0.0306	0.2708	0.8536
No log	12.96	39	3.9633	0.18	0.8805	7.0056	0.18	0.0306	0.2619	0.8535
No log	13.96	42	3.9614	0.18	0.8804	6.9981	0.18	0.0306	0.2617	0.8532
No log	14.96	45	3.9598	0.18	0.8804	6.9923	0.18	0.0306	0.2669	0.8531
No log	15.96	48	3.9586	0.18	0.8803	6.9334	0.18	0.0306	0.2669	0.8529
No log	16.96	51	3.9578	0.18	0.8802	6.9237	0.18	0.0306	0.2716	0.8522
No log	17.96	54	3.9576	0.18	0.8802	6.8704	0.18	0.0306	0.2666	0.8521
No log	18.96	57	3.9574	0.18	0.8802	6.8662	0.18	0.0306	0.2664	0.8523
No log	19.96	60	3.9568	0.18	0.8801	6.8641	0.18	0.0306	0.2614	0.8518
No log	20.96	63	3.9566	0.18	0.8801	6.8634	0.18	0.0306	0.2659	0.8516
No log	21.96	66	3.9563	0.18	0.8800	6.8632	0.18	0.0306	0.2612	0.8516
No log	22.96	69	3.9561	0.18	0.8800	6.8620	0.18	0.0306	0.2612	0.8513
No log	23.96	72	3.9561	0.18	0.8800	6.8611	0.18	0.0306	0.2612	0.8513
No log	24.96	75	3.9560	0.18	0.8800	6.8606	0.18	0.0306	0.2612	0.8512

Framework versions

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