<!-- 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-gec-combine_data
This model is a fine-tuned version of t5-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5624
- Rouge1: 76.0801
- Rouge2: 65.3291
- Rougel: 75.4097
- Rougelsum: 75.4189
- Gen Len: 16.8811
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: 1e-05
- 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: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
|---|---|---|---|---|---|---|---|---|
| 0.8413 | 0.45 | 500 | 0.6549 | 74.1413 | 62.083 | 73.4159 | 73.4206 | 16.9847 |
| 0.6945 | 0.9 | 1000 | 0.6147 | 75.0476 | 63.5645 | 74.3777 | 74.3726 | 16.926 |
| 0.6551 | 1.35 | 1500 | 0.5972 | 75.3742 | 64.1326 | 74.7249 | 74.7271 | 16.9048 |
| 0.6429 | 1.81 | 2000 | 0.5853 | 75.6 | 64.4944 | 74.9325 | 74.9335 | 16.9061 |
| 0.6142 | 2.26 | 2500 | 0.5783 | 75.7553 | 64.7554 | 75.0816 | 75.09 | 16.8953 |
| 0.6116 | 2.71 | 3000 | 0.5704 | 75.8622 | 64.9522 | 75.1886 | 75.1969 | 16.8863 |
| 0.5991 | 3.16 | 3500 | 0.5681 | 75.9079 | 65.0892 | 75.2619 | 75.2666 | 16.8887 |
| 0.5933 | 3.61 | 4000 | 0.5664 | 76.0166 | 65.2437 | 75.3563 | 75.3664 | 16.8842 |
| 0.5904 | 4.06 | 4500 | 0.5645 | 76.0398 | 65.2934 | 75.3759 | 75.3875 | 16.8849 |
| 0.5962 | 4.51 | 5000 | 0.5624 | 76.0458 | 65.2794 | 75.3762 | 75.3876 | 16.882 |
| 0.5749 | 4.96 | 5500 | 0.5624 | 76.0801 | 65.3291 | 75.4097 | 75.4189 | 16.8811 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.12.1+cu113
- Datasets 2.7.1
- Tokenizers 0.13.2
