longt5_xl_sfd_20

This model is a fine-tuned version of google/long-t5-tglobal-xl on the tau/scrolls summ_screen_fd dataset. It achieves the following results on the evaluation set:

• Loss: 4.8167

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: 0.001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 32
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• num_epochs: 20.0

Training results

Training Loss	Epoch	Step	Validation Loss
2.9377	0.97	14	2.7965
2.3247	1.95	28	2.2995
1.9943	2.99	43	2.3308
1.7056	3.97	57	2.3369
1.4471	4.94	71	2.5532
1.1932	5.98	86	2.6961
0.9199	6.96	100	2.6453
0.7571	8.0	115	3.0978
0.6287	8.97	129	3.1458
0.4939	9.95	143	3.5686
0.376	10.99	158	3.4260
0.313	11.97	172	3.9302
0.2562	12.94	186	3.7433
0.2051	13.98	201	3.9113
0.1802	14.96	215	4.0450
0.1378	16.0	230	3.9397
0.1546	16.97	244	4.3079
0.1375	17.95	258	4.5525
0.1005	18.99	273	4.9005
0.0808	19.48	280	4.8167

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.2+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2