File size: 1,578 Bytes
09249f2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ea0c9f9
09249f2
 
 
 
ea0c9f9
 
09249f2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ea0c9f9
 
 
 
09249f2
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
---
library_name: transformers
license: gemma
base_model: google/paligemma2-3b-pt-448
tags:
- generated_from_trainer
model-index:
- name: paligemma-architecture
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# paligemma-architecture

This model is a fine-tuned version of [google/paligemma2-3b-pt-448](https://huggingface.co/google/paligemma2-3b-pt-448) on a custom architecture dataset.


## Training procedure

Followed the [notebook from smol-vision](https://github.com/merveenoyan/smol-vision/blob/main/Fine_tune_PaliGemma.ipynb), adjusted dataset loading and some parameters.

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 1
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 8
- 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: 4

### Training results

TrainOutput(global_step=352, training_loss=7.797419488430023,
metrics={'train_runtime': 1653.6164, 'train_samples_per_second': 1.705,
'train_steps_per_second': 0.213, 'total_flos': 5.772661476596784e+16,
'train_loss': 7.797419488430023, 'epoch': 3.9645390070921986})

### Framework versions

- Transformers 4.50.0.dev0
- Pytorch 2.6.0+cu124
- Datasets 3.4.0
- Tokenizers 0.21.0