bcbbb5fd9e38f23f037c97024bb13257

This model is a fine-tuned version of facebook/opt-1.3b on the dim/tldr_news dataset. It achieves the following results on the evaluation set:

• Loss: 1.2295
• Data Size: 1.0
• Epoch Runtime: 36.5672
• Accuracy: 0.7180
• F1 Macro: 0.7605
• Rouge1: 0.7188
• Rouge2: 0.0
• Rougel: 0.7180
• Rougelsum: 0.7188

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	1.8950	0	2.9405	0.1619	0.1553	0.1619	0.0	0.1612	0.1619
No log	1	178	1.3171	0.0078	4.3394	0.5057	0.3456	0.5064	0.0	0.5064	0.5057
No log	2	356	1.3258	0.0156	4.8606	0.6151	0.4203	0.6165	0.0	0.6165	0.6151
No log	3	534	0.8288	0.0312	6.6482	0.7145	0.5591	0.7152	0.0	0.7145	0.7145
No log	4	712	0.7312	0.0625	8.8731	0.7237	0.5977	0.7251	0.0	0.7244	0.7230
No log	5	890	0.7379	0.125	10.6142	0.7138	0.6992	0.7145	0.0	0.7145	0.7145
0.0485	6	1068	0.6101	0.25	15.3577	0.7607	0.7473	0.7617	0.0	0.7614	0.7614
0.5181	7	1246	0.6024	0.5	24.3414	0.7635	0.7995	0.7642	0.0	0.7635	0.7635
0.4091	8.0	1424	0.8403	1.0	39.8790	0.7322	0.7220	0.7330	0.0	0.7330	0.7315
0.2411	9.0	1602	0.8390	1.0	33.3271	0.7649	0.7969	0.7649	0.0	0.7649	0.7649
0.1409	10.0	1780	1.0282	1.0	35.6351	0.7450	0.7844	0.7454	0.0	0.7457	0.7450
0.1504	11.0	1958	1.2295	1.0	36.5672	0.7180	0.7605	0.7188	0.0	0.7180	0.7188

Framework versions

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