RichardErkhov/Vexemous_-_bart-base-finetuned-xsum-8bits

Quantization made by Richard Erkhov.

Github

Discord

Request more models

bart-base-finetuned-xsum - bnb 8bits

• Model creator: https://huggingface.co/Vexemous/
• Original model: https://huggingface.co/Vexemous/bart-base-finetuned-xsum/

Original model description:

license: apache-2.0 base_model: facebook/bart-base tags: - generated_from_trainer datasets: - xsum metrics: - rouge model-index: - name: bart-base-finetuned-xsum results: - task: name: Sequence-to-sequence Language Modeling type: text2text-generation dataset: name: xsum type: xsum config: default split: train[:10%] args: default metrics: - name: Rouge1 type: rouge value: 35.8214 pipeline_tag: summarization

bart-base-finetuned-xsum

This model is a fine-tuned version of facebook/bart-base on the xsum dataset. It achieves the following results on the evaluation set:

• Loss: 1.9356
• Rouge1: 35.8214
• Rouge2: 14.7565
• Rougel: 29.4566
• Rougelsum: 29.4496
• Gen Len: 19.562

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
2.301	1.0	1148	1.9684	34.4715	13.6638	28.1147	28.1204	19.5816
2.1197	2.0	2296	1.9442	35.2502	14.284	28.8462	28.8384	19.5546
1.9804	3.0	3444	1.9406	35.7799	14.7422	29.3669	29.3742	19.5326
1.8891	4.0	4592	1.9349	35.5151	14.4668	29.0359	29.0484	19.5492
1.827	5.0	5740	1.9356	35.8214	14.7565	29.4566	29.4496	19.562

Framework versions

• Transformers 4.40.1
• Pytorch 1.13.1+cu117
• Datasets 2.19.0
• Tokenizers 0.19.1