<!-- 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-small-e-snli-generation-label_and_explanation-selected-b64
This model is a fine-tuned version of t5-small on the esnli dataset. It achieves the following results on the evaluation set:
- Loss: 1.9257
- Accuracy: 0.8733
- F1: 0.8730
- Bertscore F1: 0.9356
- Rouge1: 0.6144
- Rouge2: 0.4096
- Rougel: 0.5592
- Rougelsum: 0.5611
- Bleu: 0.4224
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: 0.001
- train_batch_size: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.05
- num_epochs: 10
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | Bertscore F1 | Rouge1 | Rouge2 | Rougel | Rougelsum | Bleu |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1.6638 | 0.23 | 2000 | 2.0039 | 0.7883 | 0.7869 | 0.9274 | 0.5705 | 0.3601 | 0.5175 | 0.5192 | 0.3730 |
| 1.2998 | 0.47 | 4000 | 1.9378 | 0.8283 | 0.8293 | 0.9303 | 0.5861 | 0.3748 | 0.5310 | 0.5329 | 0.3854 |
| 1.2351 | 0.7 | 6000 | 1.8752 | 0.8431 | 0.8437 | 0.9321 | 0.5951 | 0.3880 | 0.5411 | 0.5430 | 0.3954 |
| 1.1948 | 0.93 | 8000 | 1.9346 | 0.8536 | 0.8529 | 0.9333 | 0.6018 | 0.3931 | 0.5451 | 0.5472 | 0.4006 |
| 1.1537 | 1.16 | 10000 | 1.8881 | 0.8654 | 0.8647 | 0.9332 | 0.6070 | 0.4023 | 0.5483 | 0.5506 | 0.4096 |
| 1.1298 | 1.4 | 12000 | 1.9265 | 0.8690 | 0.8685 | 0.9337 | 0.6053 | 0.3988 | 0.5507 | 0.5526 | 0.4093 |
| 1.1219 | 1.63 | 14000 | 1.9017 | 0.8713 | 0.8714 | 0.9332 | 0.6029 | 0.3941 | 0.5470 | 0.5489 | 0.4042 |
| 1.1088 | 1.86 | 16000 | 1.9257 | 0.8733 | 0.8730 | 0.9356 | 0.6144 | 0.4096 | 0.5592 | 0.5611 | 0.4224 |
Framework versions
- Transformers 4.27.4
- Pytorch 2.0.0+cu117
- Datasets 2.11.0
- Tokenizers 0.13.2
