synthea_t5_summarization_model

This model is a fine-tuned version of t5-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.2906
• Rouge1: 0.4543
• Rouge2: 0.137
• Rougel: 0.4022
• Rougelsum: 0.4025
• Gen Len: 11.1279

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
No log	1.0	22	1.9114	0.289	0.0651	0.2709	0.2713	9.6279
No log	2.0	44	1.8289	0.3265	0.0822	0.2925	0.2935	10.2093
No log	3.0	66	1.7603	0.3617	0.107	0.3224	0.323	10.9651
No log	4.0	88	1.7081	0.3505	0.1017	0.3142	0.3146	11.7674
No log	5.0	110	1.6651	0.3497	0.0935	0.3055	0.3066	11.9419
No log	6.0	132	1.6200	0.3701	0.1022	0.3339	0.3348	11.7209
No log	7.0	154	1.5865	0.3726	0.1045	0.3328	0.3336	11.6279
No log	8.0	176	1.5552	0.3802	0.1049	0.3412	0.3417	11.4419
No log	9.0	198	1.5237	0.3982	0.115	0.3519	0.3533	11.3721
No log	10.0	220	1.4836	0.41	0.1188	0.3643	0.3645	11.4767
No log	11.0	242	1.4708	0.391	0.1142	0.3492	0.3491	11.6977
No log	12.0	264	1.4429	0.4157	0.1184	0.3689	0.3687	11.1977
No log	13.0	286	1.4312	0.4229	0.1204	0.3738	0.3741	11.0698
No log	14.0	308	1.4162	0.4231	0.1361	0.3806	0.3805	11.0465
No log	15.0	330	1.4011	0.4341	0.1406	0.3856	0.386	10.8953
No log	16.0	352	1.3877	0.439	0.1373	0.3942	0.3952	11.407
No log	17.0	374	1.3794	0.4488	0.1442	0.3987	0.3997	11.0581
No log	18.0	396	1.3673	0.4445	0.1418	0.3972	0.3979	11.186
No log	19.0	418	1.3581	0.4529	0.1375	0.4037	0.4047	11.1279
No log	20.0	440	1.3515	0.4378	0.1216	0.3921	0.3921	11.0
No log	21.0	462	1.3430	0.4533	0.1344	0.3996	0.4012	10.6512
No log	22.0	484	1.3390	0.4489	0.1426	0.4041	0.4042	10.8023
1.8003	23.0	506	1.3341	0.4444	0.1359	0.3986	0.3992	10.7674
1.8003	24.0	528	1.3266	0.4525	0.1357	0.4058	0.4059	10.9186
1.8003	25.0	550	1.3290	0.4517	0.1304	0.4024	0.4027	10.7209
1.8003	26.0	572	1.3217	0.4486	0.1405	0.402	0.402	11.4186
1.8003	27.0	594	1.3194	0.4484	0.1383	0.4004	0.401	11.1279
1.8003	28.0	616	1.3158	0.4407	0.1284	0.3946	0.395	11.4302
1.8003	29.0	638	1.3111	0.4457	0.1294	0.3974	0.397	11.2558
1.8003	30.0	660	1.3075	0.4502	0.132	0.3988	0.398	11.0581
1.8003	31.0	682	1.3045	0.4482	0.1328	0.3965	0.3963	11.0698
1.8003	32.0	704	1.3012	0.4492	0.1315	0.3978	0.3971	11.093
1.8003	33.0	726	1.2988	0.4426	0.1294	0.3922	0.3923	11.2326
1.8003	34.0	748	1.2978	0.451	0.1342	0.3992	0.3998	11.1512
1.8003	35.0	770	1.2980	0.4556	0.1386	0.4062	0.4069	11.0698
1.8003	36.0	792	1.2946	0.4578	0.1387	0.4063	0.4062	11.0581
1.8003	37.0	814	1.2921	0.4549	0.138	0.4031	0.4031	11.1047
1.8003	38.0	836	1.2910	0.4531	0.1362	0.4014	0.4017	11.1512
1.8003	39.0	858	1.2907	0.4531	0.1362	0.4014	0.4017	11.0814
1.8003	40.0	880	1.2906	0.4543	0.137	0.4022	0.4025	11.1279

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.15.2