learning_rate: 4e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 32
optimizer: Use OptimizerNames.ADAMW_HF with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 2
num_epochs: 12

Training Loss	Epoch	Step	Validation Loss
No log	0.9976	318	0.2860
0.7558	1.9984	637	0.2601
0.7558	2.9992	956	0.2659
0.1793	4.0	1275	0.2886
0.0897	4.9976	1593	0.3389
0.0897	5.9984	1912	0.4511
0.0378	6.9992	2231	0.5211
0.0187	8.0	2550	0.5399
0.0187	8.9976	2868	0.5451
0.0105	9.9984	3187	0.5496
0.0092	10.9992	3506	0.5579
0.0092	11.9718	3816	0.5581

palicoqiqi/paligemma_vqav2_2

paligemma_vqav2_2