finetune_colpali_v1_2-german_ver2-4bit

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

• Loss: 0.0559
• Model Preparation Time: 0.0099

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	Model Preparation Time
No log	0.0816	1	0.3744	0.0099
1.6073	0.8163	10	0.3027	0.0099
1.2318	1.6327	20	0.2157	0.0099
0.6498	2.4490	30	0.1428	0.0099
0.5073	3.2653	40	0.1181	0.0099
0.5106	4.0816	50	0.1069	0.0099
0.2965	4.8980	60	0.0969	0.0099
0.3175	5.7143	70	0.0922	0.0099
0.1775	6.5306	80	0.1089	0.0099
0.1966	7.3469	90	0.0649	0.0099
0.1317	8.1633	100	0.0477	0.0099
0.1287	8.9796	110	0.0503	0.0099
0.2576	9.7959	120	0.0559	0.0099

Framework versions

• Transformers 4.46.1
• Pytorch 2.3.1
• Datasets 3.1.0
• Tokenizers 0.20.1