t5_sliding_window

This model is a fine-tuned version of t5-base on the nl-quad dataset. It achieves the following results on the evaluation set:

• Loss: 2.0398
• Rougel: 41.87
• Bleu: 18.36
• Meteor: 41.06
• Bert F1: 93.86
• Qsts Mean: 38.69

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: 0.0001
• train_batch_size: 48
• eval_batch_size: 48
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 10
• label_smoothing_factor: 0.05

Training results

Training Loss	Epoch	Step	Validation Loss	Rougel	Bleu	Meteor	Bert F1	Qsts Mean
2.3550	1.0	249	2.1560	38.12	14.35	36.85	93.39	33.39
2.1237	2.0	498	2.0569	39.37	15.61	38.32	93.57	35.74
1.9450	3.0	747	2.0384	40.24	16.47	39.36	93.63	35.7
1.8462	4.0	996	2.0324	40.32	16.88	39.44	93.66	37.04
1.7684	5.0	1245	2.0337	40.95	17.87	40.13	93.75	38.23
1.6971	6.0	1494	2.0398	41.87	18.36	41.06	93.86	38.69

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2