finetune_colpali_v1_2-german_ver2-8bit
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: 0.1034
- Model Preparation Time: 0.0101
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 | 0.4123 | 0.0101 |
1.2101 | 0.8163 | 10 | 0.3859 | 0.0101 |
1.2237 | 1.6327 | 20 | 0.3079 | 0.0101 |
0.6883 | 2.4490 | 30 | 0.2464 | 0.0101 |
0.4804 | 3.2653 | 40 | 0.2011 | 0.0101 |
0.5187 | 4.0816 | 50 | 0.1792 | 0.0101 |
0.3899 | 4.8980 | 60 | 0.1563 | 0.0101 |
0.203 | 5.7143 | 70 | 0.1307 | 0.0101 |
0.1897 | 6.5306 | 80 | 0.0990 | 0.0101 |
0.3326 | 7.3469 | 90 | 0.1246 | 0.0101 |
0.3578 | 8.1633 | 100 | 0.1429 | 0.0101 |
0.3568 | 8.9796 | 110 | 0.1148 | 0.0101 |
0.1411 | 9.7959 | 120 | 0.1034 | 0.0101 |
Framework versions
- Transformers 4.46.1
- Pytorch 2.3.1
- Datasets 3.1.0
- Tokenizers 0.20.1
Model tree for svenbl80/finetune_colpali_v1_2-german_ver2-8bit
Base model
google/paligemma-3b-pt-448
Finetuned
vidore/colpaligemma-3b-pt-448-base