scofield commited on
Commit
028d935
1 Parent(s): 901d3ac

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +98 -0
README.md ADDED
@@ -0,0 +1,98 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-nc-sa-4.0
3
+ tags:
4
+ - generated_from_trainer
5
+ datasets:
6
+ - cord-layoutlmv3
7
+ metrics:
8
+ - precision
9
+ - recall
10
+ - f1
11
+ - accuracy
12
+ model-index:
13
+ - name: layoutlmv3-finetuned-cord_100
14
+ results:
15
+ - task:
16
+ name: Token Classification
17
+ type: token-classification
18
+ dataset:
19
+ name: cord-layoutlmv3
20
+ type: cord-layoutlmv3
21
+ config: cord
22
+ split: train
23
+ args: cord
24
+ metrics:
25
+ - name: Precision
26
+ type: precision
27
+ value: 0.9289415247964471
28
+ - name: Recall
29
+ type: recall
30
+ value: 0.9393712574850299
31
+ - name: F1
32
+ type: f1
33
+ value: 0.9341272794938594
34
+ - name: Accuracy
35
+ type: accuracy
36
+ value: 0.9393039049235993
37
+ ---
38
+
39
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
40
+ should probably proofread and complete it, then remove this comment. -->
41
+
42
+ # layoutlmv3-finetuned-cord_100
43
+
44
+ This model is a fine-tuned version of [microsoft/layoutlmv3-base](https://huggingface.co/microsoft/layoutlmv3-base) on the cord-layoutlmv3 dataset.
45
+ It achieves the following results on the evaluation set:
46
+ - Loss: 0.3066
47
+ - Precision: 0.9289
48
+ - Recall: 0.9394
49
+ - F1: 0.9341
50
+ - Accuracy: 0.9393
51
+
52
+ ## Model description
53
+
54
+ More information needed
55
+
56
+ ## Intended uses & limitations
57
+
58
+ More information needed
59
+
60
+ ## Training and evaluation data
61
+
62
+ More information needed
63
+
64
+ ## Training procedure
65
+
66
+ ### Training hyperparameters
67
+
68
+ The following hyperparameters were used during training:
69
+ - learning_rate: 1e-05
70
+ - train_batch_size: 5
71
+ - eval_batch_size: 5
72
+ - seed: 42
73
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
74
+ - lr_scheduler_type: linear
75
+ - training_steps: 2500
76
+
77
+ ### Training results
78
+
79
+ | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
80
+ |:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:|
81
+ | No log | 4.17 | 250 | 0.9691 | 0.7365 | 0.7867 | 0.7608 | 0.7992 |
82
+ | 1.3706 | 8.33 | 500 | 0.5325 | 0.8645 | 0.8885 | 0.8763 | 0.8858 |
83
+ | 1.3706 | 12.5 | 750 | 0.3943 | 0.8939 | 0.9139 | 0.9038 | 0.9151 |
84
+ | 0.3211 | 16.67 | 1000 | 0.3364 | 0.9209 | 0.9319 | 0.9263 | 0.9342 |
85
+ | 0.3211 | 20.83 | 1250 | 0.3217 | 0.9246 | 0.9364 | 0.9305 | 0.9346 |
86
+ | 0.1405 | 25.0 | 1500 | 0.3100 | 0.9296 | 0.9394 | 0.9345 | 0.9355 |
87
+ | 0.1405 | 29.17 | 1750 | 0.3113 | 0.9275 | 0.9386 | 0.9330 | 0.9363 |
88
+ | 0.076 | 33.33 | 2000 | 0.3183 | 0.9280 | 0.9364 | 0.9322 | 0.9351 |
89
+ | 0.076 | 37.5 | 2250 | 0.3125 | 0.9211 | 0.9356 | 0.9283 | 0.9363 |
90
+ | 0.0549 | 41.67 | 2500 | 0.3066 | 0.9289 | 0.9394 | 0.9341 | 0.9393 |
91
+
92
+
93
+ ### Framework versions
94
+
95
+ - Transformers 4.25.1
96
+ - Pytorch 1.13.0+cu116
97
+ - Datasets 2.8.0
98
+ - Tokenizers 0.13.2