distilbart-cnn-12-6-finetuned-arxiv

This model is a fine-tuned version of sshleifer/distilbart-cnn-12-6 on an unknown dataset. It achieves the following results on the evaluation set:

• eval_loss: 2.6022
• eval_rouge1: 42.624
• eval_rouge2: 14.9303
• eval_rougeL: 24.9256
• eval_rougeLsum: 37.688
• eval_gen_len: 126.656
• eval_runtime: 868.0833
• eval_samples_per_second: 1.152
• eval_steps_per_second: 0.288
• epoch: 1.0
• step: 2000

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: 4e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3
• mixed_precision_training: Native AMP

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1