vit-large-patch32-224-in21k-finetuned-galaxy10-decals
This model is a fine-tuned version of google/vit-large-patch32-224-in21k on the matthieulel/galaxy10_decals dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5281
- Accuracy: 0.8382
- Precision: 0.8372
- Recall: 0.8382
- F1: 0.8356
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: 0.0001
- train_batch_size: 128
- eval_batch_size: 128
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 512
- 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 |
| 1.8923 |
0.99 |
31 |
1.6725 |
0.4600 |
0.5537 |
0.4600 |
0.3682 |
| 1.1787 |
1.98 |
62 |
0.9949 |
0.7339 |
0.7513 |
0.7339 |
0.7095 |
| 0.9165 |
2.98 |
93 |
0.7946 |
0.7700 |
0.7694 |
0.7700 |
0.7540 |
| 0.802 |
4.0 |
125 |
0.6747 |
0.7948 |
0.7954 |
0.7948 |
0.7843 |
| 0.7074 |
4.99 |
156 |
0.6196 |
0.8117 |
0.8139 |
0.8117 |
0.8115 |
| 0.6424 |
5.98 |
187 |
0.6205 |
0.8021 |
0.8075 |
0.8021 |
0.7961 |
| 0.6309 |
6.98 |
218 |
0.5760 |
0.8117 |
0.8231 |
0.8117 |
0.8127 |
| 0.5682 |
8.0 |
250 |
0.5748 |
0.8151 |
0.8196 |
0.8151 |
0.8157 |
| 0.5981 |
8.99 |
281 |
0.5704 |
0.8213 |
0.8269 |
0.8213 |
0.8158 |
| 0.547 |
9.98 |
312 |
0.5282 |
0.8377 |
0.8352 |
0.8377 |
0.8345 |
| 0.5067 |
10.98 |
343 |
0.5281 |
0.8382 |
0.8372 |
0.8382 |
0.8356 |
| 0.5066 |
12.0 |
375 |
0.5441 |
0.8247 |
0.8286 |
0.8247 |
0.8219 |
| 0.4919 |
12.99 |
406 |
0.5580 |
0.8157 |
0.8236 |
0.8157 |
0.8155 |
| 0.4508 |
13.98 |
437 |
0.5269 |
0.8303 |
0.8331 |
0.8303 |
0.8279 |
| 0.4415 |
14.98 |
468 |
0.5399 |
0.8185 |
0.8249 |
0.8185 |
0.8203 |
| 0.4178 |
16.0 |
500 |
0.5229 |
0.8320 |
0.8358 |
0.8320 |
0.8301 |
| 0.366 |
16.99 |
531 |
0.5427 |
0.8275 |
0.8281 |
0.8275 |
0.8241 |
| 0.3706 |
17.98 |
562 |
0.5389 |
0.8241 |
0.8242 |
0.8241 |
0.8230 |
| 0.3609 |
18.98 |
593 |
0.5573 |
0.8247 |
0.8262 |
0.8247 |
0.8239 |
| 0.3443 |
20.0 |
625 |
0.5605 |
0.8320 |
0.8325 |
0.8320 |
0.8302 |
| 0.3214 |
20.99 |
656 |
0.5667 |
0.8281 |
0.8295 |
0.8281 |
0.8254 |
| 0.3262 |
21.98 |
687 |
0.5797 |
0.8236 |
0.8237 |
0.8236 |
0.8214 |
| 0.299 |
22.98 |
718 |
0.5938 |
0.8202 |
0.8225 |
0.8202 |
0.8195 |
| 0.2792 |
24.0 |
750 |
0.5909 |
0.8275 |
0.8258 |
0.8275 |
0.8251 |
| 0.2969 |
24.99 |
781 |
0.5658 |
0.8309 |
0.8319 |
0.8309 |
0.8306 |
| 0.2559 |
25.98 |
812 |
0.5936 |
0.8309 |
0.8294 |
0.8309 |
0.8294 |
| 0.2756 |
26.98 |
843 |
0.5898 |
0.8292 |
0.8295 |
0.8292 |
0.8287 |
| 0.254 |
28.0 |
875 |
0.6043 |
0.8303 |
0.8319 |
0.8303 |
0.8289 |
| 0.2674 |
28.99 |
906 |
0.5950 |
0.8371 |
0.8365 |
0.8371 |
0.8353 |
| 0.2432 |
29.76 |
930 |
0.5907 |
0.8360 |
0.8348 |
0.8360 |
0.8345 |
Framework versions
- Transformers 4.37.2
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.15.1
