metadata
license: mit
tags:
- generated_from_trainer
model-index:
- name: mnli_CollSgE
results: []
mnli_CollSgE
This model is a fine-tuned version of WillHeld/roberta-base-mnli on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.5544
- Acc: 0.8501
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: 2e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3
Training results
Training Loss | Epoch | Step | Validation Loss | Acc |
---|---|---|---|---|
0.4124 | 0.17 | 2000 | 0.4689 | 0.8311 |
0.4013 | 0.33 | 4000 | 0.4615 | 0.8321 |
0.3867 | 0.5 | 6000 | 0.4457 | 0.8369 |
0.3807 | 0.67 | 8000 | 0.4278 | 0.8422 |
0.3728 | 0.83 | 10000 | 0.4383 | 0.8404 |
0.3702 | 1.0 | 12000 | 0.4330 | 0.8437 |
0.2664 | 1.17 | 14000 | 0.5176 | 0.8424 |
0.2657 | 1.33 | 16000 | 0.5198 | 0.8433 |
0.2731 | 1.5 | 18000 | 0.4909 | 0.8403 |
0.2689 | 1.67 | 20000 | 0.4771 | 0.8448 |
0.2648 | 1.83 | 22000 | 0.4692 | 0.8482 |
0.2641 | 2.0 | 24000 | 0.4924 | 0.8503 |
0.1858 | 2.17 | 26000 | 0.5522 | 0.8469 |
0.1872 | 2.33 | 28000 | 0.5566 | 0.8487 |
0.1893 | 2.5 | 30000 | 0.5387 | 0.8475 |
0.1874 | 2.67 | 32000 | 0.5605 | 0.8491 |
0.1824 | 2.84 | 34000 | 0.5544 | 0.8501 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.13.0+cu117
- Datasets 2.7.1
- Tokenizers 0.12.1