t5-finetune-address-my

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

• Loss: 0.0522
• Exact Match: 83.8235
• Gen Len: 82.6103

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH 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: 100
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Exact Match	Gen Len
0.3462	1.0	304	0.1774	49.2647	80.9338
0.1542	2.0	608	0.0771	66.9118	82.3162
0.0756	3.0	912	0.0520	78.6765	83.4779
0.0459	4.0	1216	0.0547	79.4118	82.5294
0.0249	5.0	1520	0.0514	81.6176	82.4118
0.0183	6.0	1824	0.0514	82.3529	82.4338
0.013	7.0	2128	0.0507	81.6176	82.3897
0.036	8.0	2432	0.0524	83.0882	82.6176
0.0313	9.0	2736	0.0501	83.8235	82.5368
0.0106	10.0	3040	0.0523	82.3529	82.4632
0.0076	11.0	3344	0.0519	82.3529	82.6838
0.0029	12.0	3648	0.0522	83.8235	82.6103

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0