convnextv2-tiny-22k-224-finetuned-galaxy10-decals
This model is a fine-tuned version of facebook/convnextv2-tiny-22k-224 on the matthieulel/galaxy10_decals dataset. It achieves the following results on the evaluation set:
- Loss: 0.4373
- Accuracy: 0.8636
- Precision: 0.8625
- Recall: 0.8636
- F1: 0.8603
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: 64
- eval_batch_size: 64
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 30
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 |
|---|---|---|---|---|---|---|---|
| 1.5665 | 0.99 | 62 | 1.3996 | 0.5287 | 0.5180 | 0.5287 | 0.4897 |
| 0.8598 | 2.0 | 125 | 0.7433 | 0.7463 | 0.7490 | 0.7463 | 0.7396 |
| 0.7163 | 2.99 | 187 | 0.5703 | 0.7948 | 0.7919 | 0.7948 | 0.7863 |
| 0.5858 | 4.0 | 250 | 0.5194 | 0.8269 | 0.8292 | 0.8269 | 0.8190 |
| 0.5382 | 4.99 | 312 | 0.4936 | 0.8309 | 0.8314 | 0.8309 | 0.8302 |
| 0.5546 | 6.0 | 375 | 0.5054 | 0.8292 | 0.8366 | 0.8292 | 0.8234 |
| 0.5067 | 6.99 | 437 | 0.4817 | 0.8281 | 0.8324 | 0.8281 | 0.8278 |
| 0.4617 | 8.0 | 500 | 0.4565 | 0.8501 | 0.8545 | 0.8501 | 0.8497 |
| 0.4619 | 8.99 | 562 | 0.4382 | 0.8534 | 0.8520 | 0.8534 | 0.8498 |
| 0.4416 | 10.0 | 625 | 0.4330 | 0.8529 | 0.8505 | 0.8529 | 0.8504 |
| 0.4267 | 10.99 | 687 | 0.4274 | 0.8574 | 0.8575 | 0.8574 | 0.8566 |
| 0.3919 | 12.0 | 750 | 0.4407 | 0.8585 | 0.8604 | 0.8585 | 0.8563 |
| 0.3929 | 12.99 | 812 | 0.4373 | 0.8636 | 0.8625 | 0.8636 | 0.8603 |
| 0.3989 | 14.0 | 875 | 0.4351 | 0.8585 | 0.8602 | 0.8585 | 0.8577 |
| 0.3426 | 14.99 | 937 | 0.4476 | 0.8495 | 0.8500 | 0.8495 | 0.8484 |
| 0.361 | 16.0 | 1000 | 0.4463 | 0.8517 | 0.8505 | 0.8517 | 0.8501 |
| 0.2996 | 16.99 | 1062 | 0.4694 | 0.8596 | 0.8604 | 0.8596 | 0.8579 |
| 0.3394 | 18.0 | 1125 | 0.4494 | 0.8523 | 0.8526 | 0.8523 | 0.8517 |
| 0.3207 | 18.99 | 1187 | 0.4863 | 0.8506 | 0.8502 | 0.8506 | 0.8496 |
| 0.2993 | 20.0 | 1250 | 0.4748 | 0.8551 | 0.8516 | 0.8551 | 0.8521 |
| 0.287 | 20.99 | 1312 | 0.4980 | 0.8467 | 0.8436 | 0.8467 | 0.8434 |
| 0.3331 | 22.0 | 1375 | 0.4829 | 0.8546 | 0.8530 | 0.8546 | 0.8519 |
| 0.2852 | 22.99 | 1437 | 0.4943 | 0.8512 | 0.8520 | 0.8512 | 0.8508 |
| 0.2813 | 24.0 | 1500 | 0.4796 | 0.8574 | 0.8574 | 0.8574 | 0.8568 |
| 0.2807 | 24.99 | 1562 | 0.4811 | 0.8596 | 0.8576 | 0.8596 | 0.8576 |
| 0.2609 | 26.0 | 1625 | 0.4786 | 0.8608 | 0.8589 | 0.8608 | 0.8592 |
| 0.2571 | 26.99 | 1687 | 0.4777 | 0.8608 | 0.8605 | 0.8608 | 0.8602 |
| 0.2807 | 28.0 | 1750 | 0.4879 | 0.8596 | 0.8580 | 0.8596 | 0.8582 |
| 0.2578 | 28.99 | 1812 | 0.4829 | 0.8557 | 0.8550 | 0.8557 | 0.8549 |
| 0.2543 | 29.76 | 1860 | 0.4833 | 0.8563 | 0.8555 | 0.8563 | 0.8554 |
Framework versions
- Transformers 4.37.2
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.15.1