File size: 2,960 Bytes
b750673
 
 
 
 
 
3364de6
b750673
 
 
 
 
 
 
 
 
 
 
 
 
ba3fd10
 
 
 
 
 
40937a3
ba3fd10
 
b750673
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
abd96ec
 
b750673
 
abd96ec
b750673
 
bca0c56
b750673
 
 
 
ba3fd10
 
 
 
 
 
b750673
 
 
 
 
 
3364de6
b750673
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
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
---
license: mit
base_model: naver-clova-ix/donut-base
tags:
- generated_from_trainer
metrics:
- bleu
- wer
model-index:
- name: donut-base-sroie-metrics-combined-new
  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. -->

# donut-base-sroie-metrics-combined-new

This model is a fine-tuned version of [naver-clova-ix/donut-base](https://huggingface.co/naver-clova-ix/donut-base) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.4671
- Bleu: 0.0662
- Precisions: [0.785140562248996, 0.6825396825396826, 0.6197916666666666, 0.5626911314984709]
- Brevity Penalty: 0.1007
- Length Ratio: 0.3035
- Translation Length: 498
- Reference Length: 1641
- Cer: 0.7528
- Wer: 0.8385

## 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: 2e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 2
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 4
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step | Validation Loss | Bleu   | Precisions                                                                         | Brevity Penalty | Length Ratio | Translation Length | Reference Length | Cer    | Wer    |
|:-------------:|:-----:|:----:|:---------------:|:------:|:----------------------------------------------------------------------------------:|:---------------:|:------------:|:------------------:|:----------------:|:------:|:------:|
| 3.6559        | 1.0   | 253  | 1.5613          | 0.0007 | [0.5056179775280899, 0.1943127962085308, 0.07692307692307693, 0.02830188679245283] | 0.0058          | 0.1627       | 267                | 1641             | 0.8768 | 0.9436 |
| 1.2493        | 2.0   | 506  | 0.6697          | 0.0409 | [0.6560509554140127, 0.5048309178743962, 0.4481792717086835, 0.39]                 | 0.0834          | 0.2870       | 471                | 1641             | 0.7766 | 0.8837 |
| 0.9257        | 3.0   | 759  | 0.5168          | 0.0594 | [0.75, 0.6275862068965518, 0.5714285714285714, 0.5264797507788161]                 | 0.0968          | 0.2998       | 492                | 1641             | 0.7570 | 0.8499 |
| 0.6416        | 4.0   | 1012 | 0.4671          | 0.0662 | [0.785140562248996, 0.6825396825396826, 0.6197916666666666, 0.5626911314984709]    | 0.1007          | 0.3035       | 498                | 1641             | 0.7528 | 0.8385 |


### Framework versions

- Transformers 4.41.0.dev0
- Pytorch 2.1.0
- Datasets 2.19.1
- Tokenizers 0.19.1