<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
t5-base-DreamBank-Generation-NER-Char
This model is a fine-tuned version of t5-base on the DremBan dataset to detect which characters are present in a given report, following the Hall & Van de Castle (HVDC) framework. Please note that, during training: i) it was not specified to which features the characters were associated with; ii) in accordance with the HVDC system, the presence of the dreamer is not assessed.
It achieves the following results on the evaluation set:
- Loss: 0.4674
- Rouge1: 0.7853
- Rouge2: 0.6927
- Rougel: 0.7564
- Rougelsum: 0.7565
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: 0.0002
- 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: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 93 | 0.6486 | 0.5936 | 0.4495 | 0.5705 | 0.5701 |
| No log | 2.0 | 186 | 0.5363 | 0.7196 | 0.6020 | 0.6990 | 0.6983 |
| No log | 3.0 | 279 | 0.4391 | 0.7568 | 0.6459 | 0.7235 | 0.7244 |
| No log | 4.0 | 372 | 0.4223 | 0.7751 | 0.6748 | 0.7473 | 0.7477 |
| No log | 5.0 | 465 | 0.4266 | 0.7789 | 0.6746 | 0.7512 | 0.7522 |
| 0.6336 | 6.0 | 558 | 0.4296 | 0.7810 | 0.6790 | 0.7537 | 0.7539 |
| 0.6336 | 7.0 | 651 | 0.4400 | 0.7798 | 0.6808 | 0.7537 | 0.7543 |
| 0.6336 | 8.0 | 744 | 0.4497 | 0.7749 | 0.6821 | 0.7471 | 0.7481 |
| 0.6336 | 9.0 | 837 | 0.4661 | 0.7828 | 0.6910 | 0.7554 | 0.7563 |
| 0.6336 | 10.0 | 930 | 0.4674 | 0.7853 | 0.6927 | 0.7564 | 0.7565 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.12.1
- Datasets 2.5.1
- Tokenizers 0.12.1
Cite
If you use the model, please cite the pre-print.
@misc{https://doi.org/10.48550/arxiv.2302.14828,
doi = {10.48550/ARXIV.2302.14828},
url = {https://arxiv.org/abs/2302.14828},
author = {Bertolini, Lorenzo and Elce, Valentina and Michalak, Adriana and Bernardi, Giulio and Weeds, Julie},
keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {Automatic Scoring of Dream Reports' Emotional Content with Large Language Models},
publisher = {arXiv},
year = {2023},
copyright = {Creative Commons Attribution 4.0 International}
}
