ruT5_clickbait_title_generator

This model is a fine-tuned version of ai-forever/ruT5-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.2898
• Bleu: 5.3699
• Gen Len: 16.6005

Model description

Model for generating clickbait titles for non-clickbait

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
No log	1.0	114	2.7730	3.0092	15.861
No log	2.0	228	2.5655	3.9211	16.0397
No log	3.0	342	2.4573	4.7281	16.4218
No log	4.0	456	2.3735	5.0843	16.2531
3.5284	5.0	570	2.3490	5.7483	16.5533
3.5284	6.0	684	2.3254	5.6331	16.4963
3.5284	7.0	798	2.3095	5.6877	16.6898
3.5284	8.0	912	2.2949	5.6496	16.5583
2.6344	9.0	1026	2.2913	5.4792	16.6452
2.6344	10.0	1140	2.2898	5.3699	16.6005

Framework versions

• Transformers 4.39.3
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2