LED-Base-NSPCC

This model is a fine-tuned version of allenai/led-base-16384 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.8734
• Rouge1: 0.4910
• Rouge2: 0.2207
• Rougel: 0.2847
• Rougelsum: 0.2840

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: 0.0003
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
2.4662	0.9947	47	1.9451	0.4528	0.1809	0.2560	0.2558
1.6508	1.9894	94	1.8497	0.4889	0.2146	0.2720	0.2716
1.2549	2.9841	141	1.8268	0.4812	0.2092	0.2756	0.2753
0.9955	3.9788	188	1.8734	0.4910	0.2207	0.2847	0.2840

Framework versions

• Transformers 4.40.1
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1