mt5-base-finetuned-english

This model is a fine-tuned version of google/mt5-base on the xlsum dataset. It achieves the following results on the evaluation set:

• Loss: 3.3271
• Rouge-1: 31.7
• Rouge-2: 11.83
• Rouge-l: 26.43
• Gen Len: 18.88
• Bertscore: 74.3

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: 0.0005
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge-1	Rouge-2	Rouge-l	Gen Len	Bertscore
4.174	1.0	3125	3.5662	27.01	7.95	22.16	18.91	72.62
3.6577	2.0	6250	3.4304	28.84	9.09	23.64	18.87	73.32
3.4526	3.0	9375	3.3691	29.69	9.96	24.58	18.84	73.69
3.3091	4.0	12500	3.3368	30.38	10.32	25.1	18.9	73.9
3.2056	5.0	15625	3.3271	30.7	10.65	25.45	18.89	73.99

Framework versions

• Transformers 4.18.0
• Pytorch 1.11.0+cu113
• Datasets 2.2.0
• Tokenizers 0.12.1