aa368027f4b333bc73787cc95d1b9bec

This model is a fine-tuned version of facebook/opt-2.7b on the dim/tldr_news dataset. It achieves the following results on the evaluation set:

• Loss: 1.1044
• Data Size: 1.0
• Epoch Runtime: 65.1366
• Accuracy: 0.7159
• F1 Macro: 0.7583
• Rouge1: 0.7159
• Rouge2: 0.0
• Rougel: 0.7159
• Rougelsum: 0.7166

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	1.8162	0	4.1720	0.2102	0.1471	0.2102	0.0	0.2102	0.2102
No log	1	178	2.0909	0.0078	5.5281	0.4403	0.2888	0.4403	0.0	0.4396	0.4403
No log	2	356	0.8754	0.0156	9.4525	0.6911	0.5339	0.6918	0.0	0.6918	0.6918
No log	3	534	0.8581	0.0312	13.6582	0.6996	0.5578	0.7003	0.0	0.7003	0.6996
No log	4	712	0.9325	0.0625	19.2786	0.6974	0.5582	0.6982	0.0	0.6974	0.6967
No log	5	890	0.8856	0.125	26.7902	0.6776	0.6416	0.6783	0.0	0.6783	0.6776
0.052	6	1068	0.8216	0.25	29.2982	0.6626	0.6618	0.6626	0.0	0.6634	0.6626
0.6161	7	1246	0.6277	0.5	41.5831	0.7507	0.7952	0.7514	0.0	0.7514	0.7507
0.5626	8.0	1424	0.6768	1.0	70.0453	0.7436	0.7382	0.7443	0.0	0.7436	0.7436
0.4873	9.0	1602	0.7353	1.0	68.7981	0.7429	0.7676	0.7436	0.0	0.7433	0.7422
0.3543	10.0	1780	0.8205	1.0	61.5350	0.7507	0.7847	0.7507	0.0	0.7507	0.7514
0.2842	11.0	1958	1.1044	1.0	65.1366	0.7159	0.7583	0.7159	0.0	0.7159	0.7166

Framework versions

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