dinov2-large-finetuned-galaxy10-decals

This model is a fine-tuned version of facebook/dinov2-large on the matthieulel/galaxy10_decals dataset. It achieves the following results on the evaluation set:

  • Loss: 0.5869
  • Accuracy: 0.8737
  • Precision: 0.8722
  • Recall: 0.8737
  • F1: 0.8722

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
0.7564 0.99 62 0.6187 0.7976 0.8171 0.7976 0.7990
0.7766 2.0 125 0.6102 0.7852 0.8052 0.7852 0.7782
0.7103 2.99 187 0.5744 0.8089 0.8140 0.8089 0.8032
0.6704 4.0 250 0.6859 0.7745 0.7899 0.7745 0.7663
0.599 4.99 312 0.4729 0.8377 0.8412 0.8377 0.8359
0.565 6.0 375 0.4465 0.8517 0.8542 0.8517 0.8507
0.5576 6.99 437 0.4479 0.8484 0.8565 0.8484 0.8452
0.4966 8.0 500 0.4870 0.8388 0.8399 0.8388 0.8363
0.4667 8.99 562 0.4763 0.8444 0.8496 0.8444 0.8443
0.4264 10.0 625 0.4802 0.8377 0.8378 0.8377 0.8324
0.445 10.99 687 0.5246 0.8377 0.8383 0.8377 0.8343
0.3935 12.0 750 0.4883 0.8439 0.8519 0.8439 0.8434
0.374 12.99 812 0.4511 0.8568 0.8603 0.8568 0.8569
0.3551 14.0 875 0.5153 0.8546 0.8517 0.8546 0.8496
0.3573 14.99 937 0.4705 0.8579 0.8554 0.8579 0.8559
0.3385 16.0 1000 0.4547 0.8517 0.8535 0.8517 0.8517
0.2764 16.99 1062 0.5189 0.8529 0.8544 0.8529 0.8513
0.2895 18.0 1125 0.5393 0.8602 0.8587 0.8602 0.8586
0.2738 18.99 1187 0.5554 0.8405 0.8436 0.8405 0.8381
0.2563 20.0 1250 0.5478 0.8608 0.8573 0.8608 0.8574
0.2375 20.99 1312 0.5512 0.8664 0.8651 0.8664 0.8622
0.2599 22.0 1375 0.5317 0.8625 0.8607 0.8625 0.8599
0.2146 22.99 1437 0.5972 0.8568 0.8567 0.8568 0.8559
0.2132 24.0 1500 0.5934 0.8636 0.8617 0.8636 0.8606
0.2036 24.99 1562 0.5923 0.8664 0.8662 0.8664 0.8658
0.1971 26.0 1625 0.5839 0.8630 0.8621 0.8630 0.8621
0.1878 26.99 1687 0.5907 0.8625 0.8669 0.8625 0.8640
0.1922 28.0 1750 0.6058 0.8692 0.8684 0.8692 0.8680
0.1854 28.99 1812 0.6014 0.8670 0.8653 0.8670 0.8655
0.1688 29.76 1860 0.5869 0.8737 0.8722 0.8737 0.8722

Framework versions

  • Transformers 4.37.2
  • Pytorch 2.3.0
  • Datasets 2.19.1
  • Tokenizers 0.15.1
Downloads last month
19
Safetensors
Model size
304M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for matthieulel/dinov2-large-finetuned-galaxy10-decals

Finetuned
(10)
this model