9563d162e24ed7f4773da1db8df87cf8

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

• Loss: 0.4515
• Data Size: 1.0
• Epoch Runtime: 48.2736
• Accuracy: 0.8879
• F1 Macro: 0.8864

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	3.0045	0	4.0313	0.0587	0.0143
No log	1	499	3.0007	0.0078	4.5937	0.0917	0.0249
0.0301	2	998	2.9504	0.0156	4.8137	0.1505	0.0788
0.0539	3	1497	1.8811	0.0312	5.6681	0.4861	0.4176
0.0788	4	1996	1.1814	0.0625	7.2770	0.6610	0.6333
1.0449	5	2495	0.7842	0.125	10.2828	0.7588	0.7480
0.6773	6	2994	0.6233	0.25	15.5351	0.8014	0.7915
0.5469	7	3493	0.4708	0.5	26.4381	0.8533	0.8523
0.4168	8.0	3992	0.3957	1.0	48.1152	0.8798	0.8790
0.3013	9.0	4491	0.4000	1.0	47.5371	0.8715	0.8679
0.2003	10.0	4990	0.4332	1.0	47.7727	0.8707	0.8651
0.1682	11.0	5489	0.4611	1.0	47.9667	0.8795	0.8798
0.1775	12.0	5988	0.4515	1.0	48.2736	0.8879	0.8864

Framework versions

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