version1_3-10epochs-from_base

This model is a fine-tuned version of vidore/colpaligemma-3b-pt-448-base on the 3sara/validated_colpali_italian_documents_with_images dataset. It achieves the following results on the evaluation set:

• Loss: 0.2987

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use adamw_torch 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: 100
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0103	1	0.3507
0.1301	1.0205	100	0.2925
0.0914	2.0410	200	0.2754
0.0201	3.0923	300	0.2921
0.0912	4.1128	400	0.2941
0.0974	5.1538	500	0.2889
0.0237	6.1744	600	0.2724
0.0461	7.2154	700	0.2933
0.0043	8.2359	800	0.2905
0.0265	9.2769	900	0.2987

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.1