|
--- |
|
tags: |
|
- summarization |
|
language: |
|
- it |
|
metrics: |
|
- rouge |
|
model-index: |
|
- name: summarization_ilpost |
|
results: [] |
|
datasets: |
|
- ARTeLab/ilpost |
|
--- |
|
|
|
|
|
|
|
<!-- 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. --> |
|
|
|
# summarization_ilpost |
|
|
|
This model is a fine-tuned version of [gsarti/it5-base](https://huggingface.co/gsarti/it5-base) on IlPost dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 1.6020 |
|
- Rouge1: 33.7802 |
|
- Rouge2: 16.2953 |
|
- Rougel: 27.4797 |
|
- Rougelsum: 30.2273 |
|
- Gen Len: 45.3175 |
|
|
|
## 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: 6 |
|
- eval_batch_size: 6 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- num_epochs: 4.0 |
|
|
|
### Training results |
|
|
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.12.0.dev0 |
|
- Pytorch 1.9.1+cu102 |
|
- Datasets 1.12.1 |
|
- Tokenizers 0.10.3 |
|
|
