<!-- 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. -->
distilbert_add_GLUE_Experiment_mrpc_384
This model is a fine-tuned version of distilbert-base-uncased on the GLUE MRPC dataset. It achieves the following results on the evaluation set:
- Loss: 0.5935
- Accuracy: 0.7010
- F1: 0.8190
- Combined Score: 0.7600
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: 256
- eval_batch_size: 256
- seed: 10
- distributed_type: multi-GPU
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 50
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | Combined Score |
|---|---|---|---|---|---|---|
| 0.6355 | 1.0 | 15 | 0.6261 | 0.6838 | 0.8122 | 0.7480 |
| 0.6315 | 2.0 | 30 | 0.6294 | 0.6838 | 0.8122 | 0.7480 |
| 0.6327 | 3.0 | 45 | 0.6241 | 0.6838 | 0.8122 | 0.7480 |
| 0.6344 | 4.0 | 60 | 0.6285 | 0.6838 | 0.8122 | 0.7480 |
| 0.6328 | 5.0 | 75 | 0.6245 | 0.6838 | 0.8122 | 0.7480 |
| 0.6293 | 6.0 | 90 | 0.6245 | 0.6838 | 0.8122 | 0.7480 |
| 0.6341 | 7.0 | 105 | 0.6239 | 0.6838 | 0.8122 | 0.7480 |
| 0.6298 | 8.0 | 120 | 0.6240 | 0.6838 | 0.8122 | 0.7480 |
| 0.6304 | 9.0 | 135 | 0.6232 | 0.6838 | 0.8122 | 0.7480 |
| 0.6286 | 10.0 | 150 | 0.6196 | 0.6838 | 0.8122 | 0.7480 |
| 0.6045 | 11.0 | 165 | 0.5935 | 0.7010 | 0.8190 | 0.7600 |
| 0.5251 | 12.0 | 180 | 0.6129 | 0.6789 | 0.7849 | 0.7319 |
| 0.4395 | 13.0 | 195 | 0.6564 | 0.6912 | 0.7872 | 0.7392 |
| 0.3921 | 14.0 | 210 | 0.7059 | 0.6446 | 0.7173 | 0.6810 |
| 0.3399 | 15.0 | 225 | 0.7605 | 0.6887 | 0.7829 | 0.7358 |
| 0.3219 | 16.0 | 240 | 0.7614 | 0.6569 | 0.7328 | 0.6948 |
Framework versions
- Transformers 4.26.0
- Pytorch 1.14.0a0+410ce96
- Datasets 2.8.0
- Tokenizers 0.13.2
