bart-base-finetuned-cnn_dailymail

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

• Loss: 1.0437
• Rouge1: 25.3365
• Rouge2: 13.3508
• Rougel: 21.4401
• Rougelsum: 23.9107
• Bleu 1: 3.9737
• Bleu 2: 2.7698
• Bleu 3: 2.0856
• Meteor: 12.8165
• Lungime rezumat: 11.6837
• Lungime original: 48.7563

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: 5.6e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Bleu 1	Bleu 2	Bleu 3	Meteor	Lungime rezumat	Lungime original
1.3567	1.0	896	1.0741	25.256	13.2616	21.4201	23.8469	4.0588	2.8245	2.1231	12.7828	11.7437	48.7563
1.0881	2.0	1792	1.0609	25.1093	13.0973	21.1393	23.6685	3.943	2.7211	2.0277	12.6304	11.758	48.7563
1.0172	3.0	2688	1.0445	25.2209	13.2134	21.3199	23.8191	4.0205	2.7985	2.0994	12.7482	11.751	48.7563
0.9633	4.0	3584	1.0392	25.0763	13.145	21.1885	23.6877	3.9164	2.7134	2.043	12.6657	11.6963	48.7563
0.921	5.0	4480	1.0369	25.2214	13.3045	21.4317	23.8493	3.9533	2.7617	2.0827	12.7434	11.6727	48.7563
0.8865	6.0	5376	1.0377	25.3824	13.4543	21.4896	24.0024	3.9731	2.799	2.1298	12.9173	11.6563	48.7563
0.8576	7.0	6272	1.0347	25.1748	13.3232	21.3419	23.7755	3.925	2.7544	2.089	12.7437	11.6417	48.7563
0.8353	8.0	7168	1.0373	25.3485	13.3938	21.4843	23.9589	3.9384	2.7462	2.071	12.8098	11.6407	48.7563
0.8173	9.0	8064	1.0448	25.345	13.3389	21.4394	23.9221	3.9543	2.7587	2.0827	12.8046	11.6827	48.7563
0.8044	10.0	8960	1.0437	25.3365	13.3508	21.4401	23.9107	3.9737	2.7698	2.0856	12.8165	11.6837	48.7563

Framework versions

• Transformers 4.40.0
• Pytorch 2.2.2+cu118
• Datasets 2.19.0
• Tokenizers 0.19.1