metadata
license: gemma
base_model: google/paligemma-3b-pt-224
tags:
- generated_from_trainer
datasets:
- vq_av2
model-index:
- name: paligemma-vqa
results: []
paligemma-vqa
This model is a fine-tuned version of google/paligemma-3b-pt-224 on the vq_av2 dataset. It achieves the following results on the evaluation set:
- Loss: 0.0001
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: 0.02
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1200
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.0019 | 0.0736 | 500 | 0.0081 |
| 0.0004 | 0.1472 | 1000 | 0.0002 |
| 0.0003 | 0.2207 | 1500 | 0.0002 |
| 0.0001 | 0.2943 | 2000 | 0.0001 |
| 0.0001 | 0.3679 | 2500 | 0.0001 |
| 0.0001 | 0.4415 | 3000 | 0.0001 |
| 0.0002 | 0.5151 | 3500 | 0.0002 |
| 0.0001 | 0.5886 | 4000 | 0.0001 |
| 0.0001 | 0.6622 | 4500 | 0.0001 |
| 0.0001 | 0.7358 | 5000 | 0.0001 |
| 0.0001 | 0.8094 | 5500 | 0.0001 |
| 0.0001 | 0.8830 | 6000 | 0.0001 |
| 0.0001 | 0.9566 | 6500 | 0.0001 |
Framework versions
- Transformers 4.41.0
- Pytorch 2.2.0+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1