ryusangwon/t5-cnndm

cnn_dailymail_350_t5-base

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

• Loss: 0.8973
• Rouge1: 0.2524
• Rouge2: 0.1238
• Rougel: 0.2084
• Rougelsum: 0.2083
• Gen Len: 18.9993

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
0.8756	0.45	500	0.9285	0.2483	0.1206	0.2051	0.2051	18.9993
0.8719	0.89	1000	0.9147	0.2496	0.1221	0.2063	0.2062	18.9999
0.8407	1.34	1500	0.9101	0.2497	0.1217	0.2061	0.2061	18.9999
0.8433	1.78	2000	0.9054	0.2512	0.1225	0.2072	0.2072	18.9995
0.8346	2.23	2500	0.9048	0.2515	0.123	0.2074	0.2074	18.9998
0.8308	2.67	3000	0.9037	0.2504	0.1226	0.2073	0.2073	18.9996
0.8189	3.12	3500	0.9022	0.2517	0.1232	0.2082	0.2081	19.0
0.8275	3.57	4000	0.9011	0.2514	0.123	0.2076	0.2076	19.0
0.8272	4.01	4500	0.9010	0.2517	0.1236	0.2081	0.2081	18.9993
0.819	4.46	5000	0.8994	0.2517	0.1235	0.208	0.2079	18.999
0.8096	4.9	5500	0.9001	0.2518	0.1236	0.208	0.208	18.9992
0.823	5.35	6000	0.8976	0.2519	0.1232	0.208	0.208	18.9993
0.8205	5.8	6500	0.8979	0.2516	0.1234	0.2079	0.2079	18.9996
0.8136	6.24	7000	0.8981	0.2515	0.1232	0.2078	0.2078	18.9992
0.8117	6.69	7500	0.8984	0.2519	0.1236	0.2081	0.208	18.9996
0.8039	7.13	8000	0.8979	0.2524	0.1237	0.2083	0.2083	18.9993
0.7934	7.58	8500	0.8981	0.2517	0.1235	0.2078	0.2078	18.9992
0.7947	8.02	9000	0.8979	0.252	0.1237	0.2081	0.2081	18.9989
0.8189	8.47	9500	0.8974	0.2523	0.1237	0.2083	0.2083	18.999
0.8102	8.92	10000	0.8976	0.2523	0.1237	0.2084	0.2084	18.9991
0.8029	9.36	10500	0.8978	0.2523	0.1237	0.2083	0.2083	18.9992
0.8004	9.81	11000	0.8973	0.2524	0.1238	0.2084	0.2083	18.9993

Framework versions

• Transformers 4.36.2
• Pytorch 2.0.1+cu117
• Datasets 2.15.0
• Tokenizers 0.15.0