quickdraw-ConvNeXT-Tiny-Finetune
This model is a fine-tuned version of facebook/convnextv2-tiny-1k-224 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.9012
- Accuracy: 0.7697
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: 10000
- num_epochs: 8
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 1.2089 | 0.5688 | 5000 | 1.1779 | 0.7020 |
| 1.0859 | 1.1377 | 10000 | 1.1145 | 0.7167 |
| 1.0115 | 1.7065 | 15000 | 1.0129 | 0.7402 |
| 0.8966 | 2.2753 | 20000 | 0.9684 | 0.7533 |
| 0.8868 | 2.8441 | 25000 | 0.9375 | 0.7600 |
| 0.7743 | 3.4130 | 30000 | 0.9292 | 0.7638 |
| 0.7735 | 3.9818 | 35000 | 0.9005 | 0.7707 |
| 0.6379 | 4.5506 | 40000 | 0.9470 | 0.7675 |
| 0.4587 | 5.1195 | 45000 | 1.0663 | 0.7632 |
| 0.469 | 5.6883 | 50000 | 1.0687 | 0.7642 |
| 0.3053 | 6.2571 | 55000 | 1.2674 | 0.7561 |
| 0.3087 | 6.8259 | 60000 | 1.3039 | 0.7563 |
| 0.215 | 7.3948 | 65000 | 1.4453 | 0.7499 |
| 0.2128 | 7.9636 | 70000 | 1.4542 | 0.7500 |
Framework versions
- Transformers 4.40.2
- Pytorch 2.2.1
- Datasets 2.19.1
- Tokenizers 0.19.1