convnextv2-tiny-22k-384-finetuned-galaxy10-decals

This model is a fine-tuned version of facebook/convnextv2-tiny-22k-384 on the matthieulel/galaxy10_decals dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4641
  • Accuracy: 0.8675
  • Precision: 0.8664
  • Recall: 0.8675
  • F1: 0.8661

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
1.5875 0.99 62 1.3967 0.5423 0.5237 0.5423 0.5001
0.8561 2.0 125 0.7084 0.7773 0.7769 0.7773 0.7692
0.7139 2.99 187 0.5607 0.8230 0.8201 0.8230 0.8148
0.5799 4.0 250 0.4982 0.8410 0.8428 0.8410 0.8324
0.5352 4.99 312 0.4781 0.8461 0.8470 0.8461 0.8446
0.539 6.0 375 0.4538 0.8523 0.8578 0.8523 0.8482
0.5129 6.99 437 0.4496 0.8472 0.8486 0.8472 0.8468
0.4685 8.0 500 0.4458 0.8551 0.8589 0.8551 0.8542
0.4675 8.99 562 0.4352 0.8613 0.8651 0.8613 0.8579
0.441 10.0 625 0.4076 0.8636 0.8616 0.8636 0.8607
0.4214 10.99 687 0.4346 0.8517 0.8556 0.8517 0.8522
0.4016 12.0 750 0.4300 0.8591 0.8597 0.8591 0.8573
0.3913 12.99 812 0.4164 0.8625 0.8624 0.8625 0.8601
0.3882 14.0 875 0.4246 0.8591 0.8618 0.8591 0.8570
0.3341 14.99 937 0.4321 0.8574 0.8555 0.8574 0.8555
0.3522 16.0 1000 0.4322 0.8568 0.8561 0.8568 0.8542
0.2824 16.99 1062 0.4364 0.8608 0.8606 0.8608 0.8586
0.315 18.0 1125 0.4495 0.8579 0.8581 0.8579 0.8559
0.3084 18.99 1187 0.4536 0.8608 0.8593 0.8608 0.8590
0.2864 20.0 1250 0.4417 0.8630 0.8621 0.8630 0.8607
0.2654 20.99 1312 0.4585 0.8630 0.8628 0.8630 0.8610
0.3067 22.0 1375 0.4673 0.8557 0.8562 0.8557 0.8538
0.2771 22.99 1437 0.4679 0.8596 0.8577 0.8596 0.8577
0.2588 24.0 1500 0.4616 0.8653 0.8646 0.8653 0.8633
0.2583 24.99 1562 0.4726 0.8591 0.8572 0.8591 0.8567
0.2517 26.0 1625 0.4618 0.8630 0.8625 0.8630 0.8619
0.2454 26.99 1687 0.4612 0.8641 0.8630 0.8641 0.8629
0.259 28.0 1750 0.4685 0.8613 0.8595 0.8613 0.8594
0.2388 28.99 1812 0.4668 0.8636 0.8622 0.8636 0.8620
0.2414 29.76 1860 0.4641 0.8675 0.8664 0.8675 0.8661

Framework versions

  • Transformers 4.37.2
  • Pytorch 2.3.0
  • Datasets 2.19.1
  • Tokenizers 0.15.1
Downloads last month
15
Safetensors
Model size
27.9M 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/convnextv2-tiny-22k-384-finetuned-galaxy10-decals

Finetuned
(39)
this model