285bdb70c59c7bdd6449e33d878e3ee2

This model is a fine-tuned version of studio-ousia/luke-large-lite on the contemmcm/cls_20newsgroups dataset. It achieves the following results on the evaluation set:

• Loss: 0.4877
• Data Size: 1.0
• Epoch Runtime: 104.7482
• Accuracy: 0.8778
• F1 Macro: 0.8783

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Accuracy	F1 Macro
No log	0	0	3.0468	0	7.3774	0.0449	0.0071
No log	1	499	3.0167	0.0078	8.6666	0.0517	0.0049
0.0307	2	998	2.8602	0.0156	9.2394	0.2160	0.1260
0.0535	3	1497	1.3518	0.0312	11.5682	0.6235	0.5828
0.061	4	1996	0.8620	0.0625	14.9872	0.6978	0.6900
0.8843	5	2495	0.7058	0.125	21.5600	0.7818	0.7794
0.6179	6	2994	0.6287	0.25	33.8993	0.8097	0.8078
0.514	7	3493	0.4675	0.5	59.3060	0.8561	0.8562
0.3521	8.0	3992	0.4279	1.0	107.0577	0.8715	0.8713
0.3347	9.0	4491	0.3972	1.0	104.5544	0.8800	0.8786
0.2379	10.0	4990	0.4406	1.0	105.1482	0.8823	0.8797
0.2449	11.0	5489	0.4590	1.0	105.1835	0.8780	0.8765
0.2124	12.0	5988	3.0454	1.0	104.7418	0.0507	0.0201
0.2112	13.0	6487	0.4877	1.0	104.7482	0.8778	0.8783

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1