metadata
library_name: transformers
license: apache-2.0
base_model: google/vit-large-patch32-384
tags:
- generated_from_trainer
metrics:
- precision
- recall
- f1
model-index:
- name: segmented-augmented
results: []
segmented-augmented
This model is a fine-tuned version of google/vit-large-patch32-384 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.6965
- Precision: 0.8085
- Recall: 0.8837
- F1: 0.8444
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: 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
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 |
|---|---|---|---|---|---|---|
| 0.0857 | 1.0 | 327 | 0.4359 | 0.8176 | 0.8040 | 0.8107 |
| 0.0164 | 2.0 | 654 | 0.5654 | 0.8043 | 0.8605 | 0.8315 |
| 0.0056 | 3.0 | 981 | 0.6437 | 0.8182 | 0.8671 | 0.8419 |
| 0.002 | 4.0 | 1308 | 0.6739 | 0.8055 | 0.8804 | 0.8413 |
| 0.003 | 5.0 | 1635 | 0.6965 | 0.8085 | 0.8837 | 0.8444 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.0
- Tokenizers 0.19.1