long-t5-tglobal-base

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

• Loss: 1.9401
• Rouge1: 0.1934
• Rouge2: 0.0269
• Rougel: 0.1151

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: 8
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• total_eval_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 13

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel
1.5731	0.9996	600	1.9730	0.1342	0.0151	0.0912
1.3694	1.9996	1200	1.9623	0.1371	0.0175	0.0909
1.9561	2.9992	1800	1.9565	0.1423	0.0178	0.0928
1.0882	3.9996	2400	1.9548	0.1417	0.0186	0.0900
1.4872	4.9992	3000	1.9412	0.1581	0.0212	0.1006
1.4126	5.9988	3600	1.9486	0.1589	0.0188	0.0986
1.1634	7.0	4201	1.9464	0.1756	0.0229	0.1046
0.9541	7.9996	4801	1.9401	0.1791	0.0243	0.1078
0.9153	8.9975	5400	1.9401	0.1934	0.0269	0.1151

Framework versions

• Transformers 4.41.0
• Pytorch 2.2.0
• Datasets 2.19.1
• Tokenizers 0.19.1