t5-small-asqa-ob

This model is a fine-tuned version of google/t5-small-ssm-nq on the ASQA dataset without context (closed book). It achieves the following results on the evaluation set:

• Loss: 2.8099
• Rouge1: 0.1493
• Rouge2: 0.0837
• Rougel: 0.1272
• Rougelsum: 0.1270

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: 0.0005
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
3.8208	1.0	710	2.7856	0.1267	0.0644	0.1086	0.1084
3.0532	2.0	1420	2.6247	0.1321	0.0721	0.1145	0.1144
2.5656	3.0	2130	2.5062	0.1399	0.0773	0.1213	0.1213
2.3806	4.0	2840	2.5004	0.1431	0.0805	0.1243	0.1241
2.157	5.0	3550	2.5008	0.1455	0.0808	0.1255	0.1254
2.0458	6.0	4260	2.5313	0.1510	0.0846	0.1303	0.1301
1.914	7.0	4970	2.5298	0.1585	0.0885	0.1361	0.1358
1.7479	8.0	5680	2.5832	0.1508	0.0844	0.1292	0.1291
1.6875	9.0	6390	2.5928	0.1493	0.0834	0.1281	0.1279
1.574	10.0	7100	2.6364	0.1591	0.0885	0.1364	0.1363
1.4554	11.0	7810	2.6978	0.1513	0.0849	0.1295	0.1295
1.3909	12.0	8520	2.8099	0.1493	0.0837	0.1272	0.1270

Framework versions

• Transformers 4.23.0.dev0
• Pytorch 1.12.1+cu102
• Datasets 2.5.1
• Tokenizers 0.12.1