metadata
tags:
- generated_from_trainer
metrics:
- rouge
model-index:
- name: medium
results: []
medium
This model is a fine-tuned version of 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