dinov2-small-imagenet1k-1-layer-finetuned-galaxy10-decals
This model is a fine-tuned version of facebook/dinov2-small-imagenet1k-1-layer on the matthieulel/galaxy10_decals dataset. It achieves the following results on the evaluation set:
- Loss: 0.5373
- Accuracy: 0.8563
- Precision: 0.8536
- Recall: 0.8563
- F1: 0.8543
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 |
|---|---|---|---|---|---|---|---|
| 1.0027 | 0.99 | 62 | 0.8262 | 0.7080 | 0.7231 | 0.7080 | 0.7017 |
| 0.8374 | 2.0 | 125 | 0.6129 | 0.7948 | 0.7960 | 0.7948 | 0.7899 |
| 0.7753 | 2.99 | 187 | 0.6555 | 0.7892 | 0.7921 | 0.7892 | 0.7787 |
| 0.7165 | 4.0 | 250 | 0.5862 | 0.8005 | 0.8053 | 0.8005 | 0.7970 |
| 0.6477 | 4.99 | 312 | 0.6183 | 0.7965 | 0.8119 | 0.7965 | 0.7985 |
| 0.6892 | 6.0 | 375 | 0.5310 | 0.8247 | 0.8275 | 0.8247 | 0.8195 |
| 0.6171 | 6.99 | 437 | 0.5678 | 0.8083 | 0.8157 | 0.8083 | 0.8022 |
| 0.55 | 8.0 | 500 | 0.4961 | 0.8326 | 0.8353 | 0.8326 | 0.8316 |
| 0.5615 | 8.99 | 562 | 0.5033 | 0.8309 | 0.8312 | 0.8309 | 0.8274 |
| 0.5107 | 10.0 | 625 | 0.5162 | 0.8191 | 0.8164 | 0.8191 | 0.8152 |
| 0.5237 | 10.99 | 687 | 0.4790 | 0.8422 | 0.8452 | 0.8422 | 0.8381 |
| 0.4954 | 12.0 | 750 | 0.4782 | 0.8422 | 0.8430 | 0.8422 | 0.8373 |
| 0.4887 | 12.99 | 812 | 0.4689 | 0.8371 | 0.8395 | 0.8371 | 0.8358 |
| 0.4629 | 14.0 | 875 | 0.4541 | 0.8523 | 0.8500 | 0.8523 | 0.8502 |
| 0.4486 | 14.99 | 937 | 0.4755 | 0.8405 | 0.8400 | 0.8405 | 0.8394 |
| 0.4361 | 16.0 | 1000 | 0.4763 | 0.8371 | 0.8392 | 0.8371 | 0.8370 |
| 0.3833 | 16.99 | 1062 | 0.4982 | 0.8416 | 0.8429 | 0.8416 | 0.8396 |
| 0.3788 | 18.0 | 1125 | 0.5632 | 0.8292 | 0.8365 | 0.8292 | 0.8267 |
| 0.3722 | 18.99 | 1187 | 0.5162 | 0.8388 | 0.8364 | 0.8388 | 0.8357 |
| 0.3467 | 20.0 | 1250 | 0.5125 | 0.8399 | 0.8357 | 0.8399 | 0.8342 |
| 0.3518 | 20.99 | 1312 | 0.5569 | 0.8309 | 0.8327 | 0.8309 | 0.8276 |
| 0.3432 | 22.0 | 1375 | 0.5032 | 0.8484 | 0.8451 | 0.8484 | 0.8454 |
| 0.3067 | 22.99 | 1437 | 0.5246 | 0.8433 | 0.8462 | 0.8433 | 0.8433 |
| 0.2923 | 24.0 | 1500 | 0.5363 | 0.8467 | 0.8482 | 0.8467 | 0.8464 |
| 0.303 | 24.99 | 1562 | 0.5435 | 0.8484 | 0.8453 | 0.8484 | 0.8457 |
| 0.2523 | 26.0 | 1625 | 0.5500 | 0.8444 | 0.8422 | 0.8444 | 0.8419 |
| 0.2523 | 26.99 | 1687 | 0.5369 | 0.8529 | 0.8533 | 0.8529 | 0.8529 |
| 0.262 | 28.0 | 1750 | 0.5373 | 0.8563 | 0.8536 | 0.8563 | 0.8543 |
| 0.232 | 28.99 | 1812 | 0.5384 | 0.8529 | 0.8509 | 0.8529 | 0.8516 |
| 0.2278 | 29.76 | 1860 | 0.5429 | 0.8512 | 0.8489 | 0.8512 | 0.8495 |
Framework versions
- Transformers 4.37.2
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.15.1