519d2357e8d898705a63d1c5e8f550ce

This model is a fine-tuned version of google-bert/bert-base-multilingual-cased on the dim/tldr_news dataset. It achieves the following results on the evaluation set:

• Loss: 1.0897
• Data Size: 1.0
• Epoch Runtime: 10.5803
• Accuracy: 0.7536
• F1 Macro: 0.7870
• Rouge1: 0.7543
• Rouge2: 0.0
• Rougel: 0.7543
• Rougelsum: 0.7528

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	Rouge1	Rouge2	Rougel	Rougelsum
No log	0	0	1.5950	0	1.2032	0.2180	0.1089	0.2180	0.0	0.2173	0.2173
No log	1	178	1.5872	0.0078	2.2795	0.2805	0.0971	0.2798	0.0	0.2798	0.2805
No log	2	356	1.5153	0.0156	1.6419	0.2401	0.0774	0.2401	0.0	0.2408	0.2393
No log	3	534	1.2839	0.0312	1.9458	0.4425	0.3121	0.4418	0.0	0.4425	0.4418
No log	4	712	0.8473	0.0625	2.4654	0.6939	0.5406	0.6946	0.0	0.6946	0.6946
No log	5	890	0.8357	0.125	3.1686	0.6875	0.5192	0.6886	0.0	0.6882	0.6882
0.0617	6	1068	0.7073	0.25	4.3275	0.7209	0.5713	0.7216	0.0	0.7209	0.7209
0.6091	7	1246	0.7171	0.5	6.5975	0.7401	0.7170	0.7415	0.0	0.7408	0.7401
0.5305	8.0	1424	0.6243	1.0	11.2915	0.7457	0.7534	0.7457	0.0	0.7464	0.7457
0.409	9.0	1602	0.6616	1.0	11.4559	0.7599	0.7910	0.7599	0.0	0.7603	0.7607
0.2709	10.0	1780	0.8822	1.0	11.0289	0.7443	0.7844	0.7450	0.0	0.7443	0.7443
0.2172	11.0	1958	0.8376	1.0	10.6638	0.7514	0.7854	0.7521	0.0	0.7521	0.7514
0.1571	12.0	2136	1.0897	1.0	10.5803	0.7536	0.7870	0.7543	0.0	0.7543	0.7528

Framework versions

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