Edit model card

my_awesome_billsum_model

This model is a fine-tuned version of t5-small on the billsum dataset. It achieves the following results on the evaluation set:

  • Loss: 2.9403
  • Rouge1: 0.1117
  • Rouge2: 0.0199
  • Rougel: 0.0955
  • Rougelsum: 0.0951
  • Gen Len: 19.0

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: 8
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 20

Training results

Training Loss Epoch Step Validation Loss Rouge1 Rouge2 Rougel Rougelsum Gen Len
No log 1.0 10 4.4002 0.1333 0.0378 0.1094 0.109 19.0
No log 2.0 20 3.8225 0.1325 0.0351 0.1085 0.1081 19.0
No log 3.0 30 3.5343 0.1343 0.0361 0.1109 0.1109 19.0
No log 4.0 40 3.3920 0.1253 0.0307 0.1069 0.1067 19.0
No log 5.0 50 3.2849 0.1239 0.0275 0.1028 0.103 19.0
No log 6.0 60 3.2041 0.1227 0.0237 0.1015 0.1016 19.0
No log 7.0 70 3.1439 0.1234 0.0218 0.1022 0.1023 19.0
No log 8.0 80 3.0979 0.1286 0.026 0.1057 0.106 19.0
No log 9.0 90 3.0624 0.1298 0.0289 0.1048 0.105 19.0
No log 10.0 100 3.0351 0.1286 0.0299 0.105 0.1053 19.0
No log 11.0 110 3.0135 0.1292 0.0288 0.1066 0.1068 19.0
No log 12.0 120 2.9956 0.1148 0.0195 0.0942 0.0938 19.0
No log 13.0 130 2.9813 0.1167 0.0195 0.0943 0.0939 19.0
No log 14.0 140 2.9697 0.1129 0.0204 0.0935 0.093 19.0
No log 15.0 150 2.9606 0.1129 0.0204 0.0935 0.093 19.0
No log 16.0 160 2.9534 0.1125 0.0198 0.0934 0.0931 19.0
No log 17.0 170 2.9478 0.1117 0.0199 0.0955 0.0951 19.0
No log 18.0 180 2.9436 0.1117 0.0199 0.0955 0.0951 19.0
No log 19.0 190 2.9411 0.1117 0.0199 0.0955 0.0951 19.0
No log 20.0 200 2.9403 0.1117 0.0199 0.0955 0.0951 19.0

Framework versions

  • Transformers 4.35.2
  • Pytorch 2.2.0.dev20231123
  • Datasets 2.15.0
  • Tokenizers 0.15.0
Downloads last month
9
Safetensors
Model size
60.5M params
Tensor type
F32
·

Finetuned from

Dataset used to train KRayRay/my_awesome_billsum_model

Evaluation results