Hugging Face's logo

TomasFAV
/
Pix2StructCzechInvoiceV0

Image-Text-to-Text
Transformers
TensorBoard
Safetensors
pix2struct
Generated from Trainer
invoice-processing
information-extraction
czech-language
document-ai
multimodal-model
generative-model
synthetic-data
Model card Files
xet
 Metrics Community
TomasFAV commited on
Commit
fc6fb86
·
verified ·
1 Parent(s): 3e160eb

Model save

Browse files
Files changed (2) hide show
  1. README.md +25 -11
  2. model.safetensors +1 -1
README.md CHANGED
@@ -4,6 +4,8 @@ license: apache-2.0
4
  base_model: google/pix2struct-docvqa-base
5
  tags:
6
  - generated_from_trainer
 
 
7
  model-index:
8
  - name: Pix2StructCzechInvoice
9
  results: []
@@ -16,12 +18,8 @@ should probably proofread and complete it, then remove this comment. -->
16
 
17
  This model is a fine-tuned version of [google/pix2struct-docvqa-base](https://huggingface.co/google/pix2struct-docvqa-base) on an unknown dataset.
18
  It achieves the following results on the evaluation set:
19
- - eval_loss: 0.2568
20
- - eval_runtime: 37.2601
21
- - eval_samples_per_second: 1.61
22
- - eval_steps_per_second: 1.61
23
- - epoch: 8.0
24
- - step: 2400
25
 
26
  ## Model description
27
 
@@ -41,18 +39,34 @@ More information needed
41
 
42
  The following hyperparameters were used during training:
43
  - learning_rate: 0.0001
44
- - train_batch_size: 2
45
  - eval_batch_size: 1
46
  - seed: 42
47
  - optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
48
  - lr_scheduler_type: cosine_with_restarts
49
- - lr_scheduler_warmup_ratio: 0.1
50
- - num_epochs: 30
51
  - mixed_precision_training: Native AMP
52
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
53
  ### Framework versions
54
 
55
- - Transformers 4.57.6
56
- - Pytorch 2.9.0+cu126
57
  - Datasets 4.0.0
58
  - Tokenizers 0.22.2
 
4
  base_model: google/pix2struct-docvqa-base
5
  tags:
6
  - generated_from_trainer
7
+ metrics:
8
+ - f1
9
  model-index:
10
  - name: Pix2StructCzechInvoice
11
  results: []
 
18
 
19
  This model is a fine-tuned version of [google/pix2struct-docvqa-base](https://huggingface.co/google/pix2struct-docvqa-base) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
+ - Loss: 0.5022
22
+ - F1: 0.4665
 
 
 
 
23
 
24
  ## Model description
25
 
 
39
 
40
  The following hyperparameters were used during training:
41
  - learning_rate: 0.0001
42
+ - train_batch_size: 4
43
  - eval_batch_size: 1
44
  - seed: 42
45
  - optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
46
  - lr_scheduler_type: cosine_with_restarts
47
+ - lr_scheduler_warmup_steps: 0.1
48
+ - num_epochs: 10
49
  - mixed_precision_training: Native AMP
50
 
51
+ ### Training results
52
+
53
+ | Training Loss | Epoch | Step | Validation Loss | F1 |
54
+ |:-------------:|:-----:|:----:|:---------------:|:------:|
55
+ | 3.1072 | 1.0 | 300 | 2.9769 | 0.0 |
56
+ | 2.6572 | 2.0 | 600 | 2.8684 | 0.0 |
57
+ | 2.4810 | 3.0 | 900 | 2.6349 | 0.0 |
58
+ | 1.7941 | 4.0 | 1200 | 1.6395 | 0.0 |
59
+ | 0.8458 | 5.0 | 1500 | 1.0680 | 0.2173 |
60
+ | 0.6198 | 6.0 | 1800 | 0.7713 | 0.4835 |
61
+ | 0.1999 | 7.0 | 2100 | 0.4331 | 0.5700 |
62
+ | 0.0946 | 8.0 | 2400 | 0.3844 | 0.5907 |
63
+ | 0.1020 | 9.0 | 2700 | 0.4066 | 0.4294 |
64
+ | 0.0842 | 10.0 | 3000 | 0.5022 | 0.4665 |
65
+
66
+
67
  ### Framework versions
68
 
69
+ - Transformers 5.0.0
70
+ - Pytorch 2.10.0+cu128
71
  - Datasets 4.0.0
72
  - Tokenizers 0.22.2
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7a505692036ffc09c6bf6be00a2df55aeca5ab273202fafa7563adb9fb3456eb
3
  size 1130333048
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2a3c2f1208358a480413d1d163bccb82fc7f3f8bef46f06844ddfdf73d93590
3
  size 1130333048