convnext-large-224-22k-1k-FV2-finetuned-memes
This model is a fine-tuned version of facebook/convnext-large-224-22k-1k on the imagefolder dataset. It achieves the following results on the evaluation set:
- Loss: 0.4290
- Accuracy: 0.8663
- Precision: 0.8617
- Recall: 0.8663
- F1: 0.8629
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: 0.00012
- 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: 10
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 |
|---|---|---|---|---|---|---|---|
| 0.8992 | 0.99 | 20 | 0.6455 | 0.7658 | 0.7512 | 0.7658 | 0.7534 |
| 0.4245 | 1.99 | 40 | 0.4008 | 0.8539 | 0.8680 | 0.8539 | 0.8541 |
| 0.2054 | 2.99 | 60 | 0.3245 | 0.8694 | 0.8631 | 0.8694 | 0.8650 |
| 0.1102 | 3.99 | 80 | 0.3231 | 0.8671 | 0.8624 | 0.8671 | 0.8645 |
| 0.0765 | 4.99 | 100 | 0.3882 | 0.8563 | 0.8603 | 0.8563 | 0.8556 |
| 0.0642 | 5.99 | 120 | 0.4133 | 0.8601 | 0.8604 | 0.8601 | 0.8598 |
| 0.0574 | 6.99 | 140 | 0.3889 | 0.8694 | 0.8657 | 0.8694 | 0.8667 |
| 0.0526 | 7.99 | 160 | 0.4145 | 0.8655 | 0.8705 | 0.8655 | 0.8670 |
| 0.0468 | 8.99 | 180 | 0.4256 | 0.8679 | 0.8642 | 0.8679 | 0.8650 |
| 0.0472 | 9.99 | 200 | 0.4290 | 0.8663 | 0.8617 | 0.8663 | 0.8629 |
Framework versions
- Transformers 4.24.0.dev0
- Pytorch 1.11.0+cu102
- Datasets 2.6.1.dev0
- Tokenizers 0.13.1