t5-small-samsum
This model is a fine-tuned version of google-t5/t5-small on an samsum dataset. It achieves the following results on the evaluation set:
- Loss: 1.6507
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 8
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 64
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| No log | 1.0 | 460 | 1.9598 |
| 2.4944 | 2.0 | 921 | 1.8661 |
| 2.0902 | 3.0 | 1381 | 1.8210 |
| 2.0173 | 4.0 | 1842 | 1.8009 |
| 1.9623 | 5.0 | 2302 | 1.7787 |
| 1.9331 | 6.0 | 2763 | 1.7637 |
| 1.903 | 7.0 | 3223 | 1.7514 |
| 1.881 | 8.0 | 3684 | 1.7390 |
| 1.8648 | 9.0 | 4144 | 1.7350 |
| 1.8463 | 10.0 | 4605 | 1.7242 |
| 1.8302 | 11.0 | 5065 | 1.7189 |
| 1.8119 | 12.0 | 5526 | 1.7098 |
| 1.8119 | 13.0 | 5986 | 1.7076 |
| 1.8007 | 14.0 | 6447 | 1.7057 |
| 1.7903 | 15.0 | 6907 | 1.6984 |
| 1.778 | 16.0 | 7368 | 1.6944 |
| 1.7639 | 17.0 | 7828 | 1.6907 |
| 1.7596 | 18.0 | 8289 | 1.6896 |
| 1.746 | 19.0 | 8749 | 1.6861 |
| 1.7342 | 20.0 | 9210 | 1.6860 |
| 1.732 | 21.0 | 9670 | 1.6808 |
| 1.719 | 22.0 | 10131 | 1.6760 |
| 1.7152 | 23.0 | 10591 | 1.6778 |
| 1.7082 | 24.0 | 11052 | 1.6762 |
| 1.7003 | 25.0 | 11512 | 1.6707 |
| 1.7003 | 26.0 | 11973 | 1.6722 |
| 1.6952 | 27.0 | 12433 | 1.6701 |
| 1.6848 | 28.0 | 12894 | 1.6671 |
| 1.6814 | 29.0 | 13354 | 1.6668 |
| 1.6743 | 30.0 | 13815 | 1.6637 |
| 1.6742 | 31.0 | 14275 | 1.6640 |
| 1.6652 | 32.0 | 14736 | 1.6624 |
| 1.6582 | 33.0 | 15196 | 1.6606 |
| 1.6575 | 34.0 | 15657 | 1.6605 |
| 1.6499 | 35.0 | 16117 | 1.6617 |
| 1.6455 | 36.0 | 16578 | 1.6601 |
| 1.6506 | 37.0 | 17038 | 1.6594 |
| 1.6506 | 38.0 | 17499 | 1.6556 |
| 1.637 | 39.0 | 17959 | 1.6570 |
| 1.6374 | 40.0 | 18420 | 1.6558 |
| 1.6303 | 41.0 | 18880 | 1.6557 |
| 1.6311 | 42.0 | 19341 | 1.6553 |
| 1.6234 | 43.0 | 19801 | 1.6570 |
| 1.619 | 44.0 | 20262 | 1.6537 |
| 1.6214 | 45.0 | 20722 | 1.6529 |
| 1.6183 | 46.0 | 21183 | 1.6542 |
| 1.609 | 47.0 | 21643 | 1.6543 |
| 1.6159 | 48.0 | 22104 | 1.6530 |
| 1.6101 | 49.0 | 22564 | 1.6524 |
| 1.6083 | 50.0 | 23025 | 1.6515 |
| 1.6083 | 51.0 | 23485 | 1.6528 |
| 1.605 | 52.0 | 23946 | 1.6526 |
| 1.6011 | 53.0 | 24406 | 1.6515 |
| 1.6028 | 54.0 | 24867 | 1.6517 |
| 1.6015 | 55.0 | 25327 | 1.6512 |
| 1.601 | 56.0 | 25788 | 1.6504 |
| 1.6007 | 57.0 | 26248 | 1.6513 |
| 1.5948 | 58.0 | 26709 | 1.6511 |
| 1.5973 | 59.0 | 27169 | 1.6515 |
| 1.5929 | 60.0 | 27630 | 1.6514 |
| 1.5955 | 61.0 | 28090 | 1.6507 |
| 1.5931 | 62.0 | 28551 | 1.6507 |
| 1.5939 | 63.0 | 29011 | 1.6507 |
| 1.5939 | 63.93 | 29440 | 1.6507 |
Framework versions
- Transformers 4.39.1
- Pytorch 2.2.1
- Datasets 2.18.0
- Tokenizers 0.15.2