6f7f30af9a3918092d173d98eb77a486

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

• Loss: 1.1040
• Data Size: 1.0
• Epoch Runtime: 36.8738
• Accuracy: 0.7443
• F1 Macro: 0.7917
• Rouge1: 0.7450
• Rouge2: 0.0
• Rougel: 0.7443
• Rougelsum: 0.7447

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	1.5962	0	2.6735	0.2599	0.1001	0.2599	0.0	0.2599	0.2592
No log	1	178	1.5405	0.0078	4.3234	0.3068	0.1548	0.3068	0.0	0.3075	0.3061
No log	2	356	1.5057	0.0156	3.8924	0.4318	0.2290	0.4318	0.0	0.4325	0.4318
No log	3	534	1.1937	0.0312	5.8330	0.4865	0.3042	0.4865	0.0	0.4858	0.4865
No log	4	712	1.0637	0.0625	7.5978	0.6300	0.4908	0.6300	0.0	0.6300	0.6307
No log	5	890	0.9300	0.125	10.3168	0.6534	0.4633	0.6541	0.0	0.6548	0.6534
0.0685	6	1068	0.7826	0.25	15.0135	0.7031	0.5703	0.7038	0.0	0.7031	0.7038
0.7148	7	1246	0.6820	0.5	23.1720	0.7443	0.5830	0.7457	0.0	0.7450	0.7443
0.553	8.0	1424	0.6457	1.0	38.9293	0.7578	0.7506	0.7578	0.0	0.7578	0.7578
0.4288	9.0	1602	0.6601	1.0	37.3702	0.7713	0.7710	0.7720	0.0	0.7720	0.7706
0.2815	10.0	1780	0.7299	1.0	37.0299	0.7507	0.7925	0.7507	0.0	0.7507	0.7507
0.1955	11.0	1958	0.9711	1.0	36.8332	0.7692	0.8077	0.7699	0.0	0.7692	0.7692
0.1563	12.0	2136	1.1040	1.0	36.8738	0.7443	0.7917	0.7450	0.0	0.7443	0.7447

Framework versions

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