Edit model card

flan-t5-large-extraction-cnndm_10000-all

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

  • Loss: 1.7044
  • Rouge1: 34.8618
  • Rouge2: 15.5978
  • Rougel: 29.7948
  • Rougelsum: 29.7581
  • 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: 24
  • seed: 1799
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 5

Training results

Training Loss Epoch Step Validation Loss Rouge1 Rouge2 Rougel Rougelsum Gen Len
2.1668 0.16 200 1.8280 33.7941 14.3114 28.7743 28.7968 19.0
1.9736 0.32 400 1.7818 34.8351 15.5548 29.8974 29.8557 18.99
1.904 0.48 600 1.7513 35.465 15.8566 30.7139 30.6596 18.986
1.8938 0.64 800 1.7440 34.6193 15.5473 30.0661 30.0019 18.99
1.8471 0.8 1000 1.7366 34.553 15.2214 29.8807 29.8419 18.99
1.8621 0.96 1200 1.7486 34.9309 15.1932 29.8973 29.8774 18.99
1.8082 1.12 1400 1.7311 35.3395 16.0976 30.2748 30.293 18.99
1.7448 1.28 1600 1.7155 35.1387 15.7462 29.924 29.9287 18.99
1.7655 1.44 1800 1.7239 35.3603 15.6355 30.3944 30.3766 19.0
1.7283 1.6 2000 1.7132 34.7368 15.4073 29.9027 29.8971 19.0
1.7463 1.76 2200 1.7171 35.0545 15.726 30.0364 30.0056 19.0
1.7462 1.92 2400 1.7044 34.8618 15.5978 29.7948 29.7581 19.0
1.719 2.08 2600 1.7285 34.9598 15.5237 29.5593 29.5803 19.0
1.6828 2.24 2800 1.7179 35.0944 15.7333 29.8381 29.7784 19.0
1.7 2.4 3000 1.7047 35.1766 15.7758 29.818 29.7859 19.0

Framework versions

  • Transformers 4.18.0
  • Pytorch 1.10.0+cu111
  • Datasets 2.5.1
  • Tokenizers 0.12.1
Downloads last month
2
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.