longt5_xl_sfd_40

This model was trained from scratch on the scrolls dataset. It achieves the following results on the evaluation set:

• Loss: 5.3992

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
0.0878	0.97	14	4.5418
0.1058	1.95	28	4.8817
0.0493	2.99	43	5.1348
0.1315	3.97	57	5.5055
0.0384	4.94	71	5.3415
0.041	5.98	86	5.1510
0.0439	6.96	100	5.1317
0.0423	8.0	115	4.8478
0.0421	8.97	129	5.3170
0.0407	9.95	143	5.0650
0.0391	10.99	158	5.6211
0.0395	11.97	172	5.3323
0.0412	12.94	186	5.4325
0.0632	13.98	201	5.3180
0.0421	14.96	215	5.1413
0.0381	16.0	230	5.5675
0.0422	16.97	244	5.5748
0.0901	17.95	258	4.2179
0.0364	18.99	273	5.6566
0.0311	19.48	280	5.3992

Framework versions

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