quickdraw-ConvNeXT-Tiny-Finetune
This model is a fine-tuned version of facebook/convnext-tiny-224 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.8413
- Accuracy: 0.7826
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: 0.0008
- train_batch_size: 512
- eval_batch_size: 512
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 5000
- num_epochs: 7
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 1.0895 | 0.5688 | 5000 | 1.0618 | 0.7278 |
| 0.8912 | 1.1377 | 10000 | 0.9315 | 0.7602 |
| 0.8658 | 1.7065 | 15000 | 0.8871 | 0.7701 |
| 0.7558 | 2.2753 | 20000 | 0.8653 | 0.7780 |
| 0.7578 | 2.8441 | 25000 | 0.8392 | 0.7826 |
| 0.6249 | 3.4130 | 30000 | 0.8584 | 0.7842 |
| 0.6243 | 3.9818 | 35000 | 0.8410 | 0.7882 |
| 0.4585 | 4.5506 | 40000 | 0.9402 | 0.7828 |
| 0.2934 | 5.1195 | 45000 | 1.0587 | 0.7758 |
| 0.2869 | 5.6883 | 50000 | 1.1280 | 0.7745 |
| 0.1837 | 6.2571 | 55000 | 1.2402 | 0.7684 |
| 0.1858 | 6.8259 | 60000 | 1.2570 | 0.7679 |
Framework versions
- Transformers 4.40.2
- Pytorch 2.2.1
- Datasets 2.19.1
- Tokenizers 0.19.1