| tags: | |
| - generated_from_trainer | |
| metrics: | |
| - rouge | |
| model-index: | |
| - name: medium | |
| results: [] | |
| <!-- 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. --> | |
| # medium | |
| This model is a fine-tuned version of [prithivida/parrot_paraphraser_on_T5](https://huggingface.co/prithivida/parrot_paraphraser_on_T5) on an unknown dataset. | |
| It achieves the following results on the evaluation set: | |
| - Loss: 0.6025 | |
| - Rouge1: 81.6007 | |
| - Rouge2: 75.1196 | |
| - Rougel: 81.4213 | |
| - Rougelsum: 81.4956 | |
| - Gen Len: 32.4286 | |
| ## 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: 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: 3.0 | |
| ### Training results | |
| | Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len | | |
| |:-------------:|:-----:|:----:|:---------------:|:-------:|:-------:|:-------:|:---------:|:-------:| | |
| | No log | 1.0 | 63 | 0.5775 | 65.0748 | 58.8985 | 64.5731 | 63.6249 | 19.0 | | |
| | No log | 2.0 | 126 | 0.5806 | 74.3055 | 69.2025 | 73.4922 | 73.0941 | 17.8571 | | |
| | No log | 3.0 | 189 | 0.6025 | 71.3808 | 66.0359 | 70.1235 | 69.4614 | 18.0 | | |
| ### Framework versions | |
| - Transformers 4.15.0 | |
| - Pytorch 1.10.1+cu113 | |
| - Datasets 1.17.0 | |
| - Tokenizers 0.10.3 | |