paligemma_vqa_lower

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: 0.0122

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• total_eval_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
4.837	0.1471	500	3.7992
0.1673	0.2943	1000	0.1149
0.0227	0.4414	1500	0.0198
0.0146	0.5886	2000	0.0138
0.0135	0.7357	2500	0.0125
0.013	0.8829	3000	0.0122

Framework versions

• Transformers 4.41.0
• Pytorch 2.2.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1