ddfcd73bbe6123a96753ab5b4f63585e

This model is a fine-tuned version of deepseek-ai/DeepSeek-R1-Distill-Qwen-7B on the dim/tldr_news dataset. It achieves the following results on the evaluation set:

• Loss: 8.8694
• Data Size: 1.0
• Epoch Runtime: 318.5639
• Accuracy: 0.7592
• F1 Macro: 0.7855
• Rouge1: 0.7599
• Rouge2: 0.0
• Rougel: 0.7592
• Rougelsum: 0.7592

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	14.8617	0	7.6791	0.3011	0.1662	0.3018	0.0	0.3011	0.3011
No log	1	178	21.2796	0.0078	9.5885	0.2891	0.1516	0.2898	0.0	0.2891	0.2891
No log	2	356	8.3821	0.0156	23.9192	0.5938	0.4083	0.5945	0.0	0.5938	0.5945
No log	3	534	5.2765	0.0312	42.1691	0.6271	0.4965	0.6278	0.0	0.6271	0.6264
No log	4	712	3.3509	0.0625	67.5970	0.6740	0.5494	0.6747	0.0	0.6733	0.6733
No log	5	890	2.9493	0.125	91.1434	0.7251	0.6998	0.7251	0.0	0.7259	0.7259
0.2973	6	1068	2.5491	0.25	141.9162	0.7756	0.7789	0.7770	0.0	0.7763	0.7763
2.0618	7	1246	2.4829	0.5	198.2432	0.7841	0.8192	0.7841	0.0	0.7841	0.7848
1.3653	8.0	1424	2.7490	1.0	331.8644	0.7557	0.7698	0.7557	0.0	0.7557	0.7557
0.5211	9.0	1602	4.4314	1.0	318.8436	0.7621	0.8016	0.7635	0.0	0.7621	0.7628
0.4182	10.0	1780	7.0370	1.0	319.3421	0.7422	0.7869	0.7429	0.0	0.7422	0.7415
0.3976	11.0	1958	8.8694	1.0	318.5639	0.7592	0.7855	0.7599	0.0	0.7592	0.7592

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.1