lilt-en-funsd
This model is a fine-tuned version of SCUT-DLVCLab/lilt-roberta-en-base on the funsd-layoutlmv3 dataset. It achieves the following results on the evaluation set:
- Loss: 1.4726
- Answer: {'precision': 0.8964677222898904, 'recall': 0.9008567931456548, 'f1': 0.8986568986568988, 'number': 817}
- Header: {'precision': 0.7446808510638298, 'recall': 0.5882352941176471, 'f1': 0.6572769953051643, 'number': 119}
- Question: {'precision': 0.8958517210944396, 'recall': 0.9424326833797586, 'f1': 0.918552036199095, 'number': 1077}
- Overall Precision: 0.8892
- Overall Recall: 0.9046
- Overall F1: 0.8968
- Overall Accuracy: 0.8387
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: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- training_steps: 2500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Answer | Header | Question | Overall Precision | Overall Recall | Overall F1 | Overall Accuracy |
|---|---|---|---|---|---|---|---|---|---|---|
| 0.4172 | 10.53 | 200 | 0.8947 | {'precision': 0.8194444444444444, 'recall': 0.8665850673194615, 'f1': 0.842355740630577, 'number': 817} | {'precision': 0.5284552845528455, 'recall': 0.5462184873949579, 'f1': 0.5371900826446281, 'number': 119} | {'precision': 0.845414847161572, 'recall': 0.8987929433611885, 'f1': 0.8712871287128714, 'number': 1077} | 0.8166 | 0.8649 | 0.8400 | 0.8019 |
| 0.0368 | 21.05 | 400 | 1.1681 | {'precision': 0.8507972665148064, 'recall': 0.9143206854345165, 'f1': 0.8814159292035397, 'number': 817} | {'precision': 0.45962732919254656, 'recall': 0.6218487394957983, 'f1': 0.5285714285714286, 'number': 119} | {'precision': 0.888671875, 'recall': 0.8449396471680595, 'f1': 0.866254164683484, 'number': 1077} | 0.8391 | 0.8599 | 0.8494 | 0.8104 |
| 0.0132 | 31.58 | 600 | 1.3663 | {'precision': 0.8438914027149321, 'recall': 0.9130966952264382, 'f1': 0.8771310993533216, 'number': 817} | {'precision': 0.6511627906976745, 'recall': 0.47058823529411764, 'f1': 0.5463414634146342, 'number': 119} | {'precision': 0.8687943262411347, 'recall': 0.9099350046425255, 'f1': 0.888888888888889, 'number': 1077} | 0.8494 | 0.8852 | 0.8669 | 0.8101 |
| 0.0061 | 42.11 | 800 | 1.4360 | {'precision': 0.8648018648018648, 'recall': 0.9082007343941249, 'f1': 0.8859701492537313, 'number': 817} | {'precision': 0.6867469879518072, 'recall': 0.4789915966386555, 'f1': 0.5643564356435644, 'number': 119} | {'precision': 0.8886910062333037, 'recall': 0.9266480965645311, 'f1': 0.9072727272727273, 'number': 1077} | 0.8706 | 0.8927 | 0.8815 | 0.8045 |
| 0.0043 | 52.63 | 1000 | 1.4084 | {'precision': 0.8550057537399309, 'recall': 0.9094247246022031, 'f1': 0.8813760379596678, 'number': 817} | {'precision': 0.6344086021505376, 'recall': 0.4957983193277311, 'f1': 0.5566037735849056, 'number': 119} | {'precision': 0.8842010771992819, 'recall': 0.914577530176416, 'f1': 0.8991328160657235, 'number': 1077} | 0.8608 | 0.8877 | 0.8741 | 0.8265 |
| 0.002 | 63.16 | 1200 | 1.4017 | {'precision': 0.8716136631330977, 'recall': 0.9057527539779682, 'f1': 0.8883553421368547, 'number': 817} | {'precision': 0.6593406593406593, 'recall': 0.5042016806722689, 'f1': 0.5714285714285715, 'number': 119} | {'precision': 0.8825088339222615, 'recall': 0.9275766016713092, 'f1': 0.9044816659121775, 'number': 1077} | 0.8682 | 0.8937 | 0.8808 | 0.8194 |
| 0.0018 | 73.68 | 1400 | 1.4379 | {'precision': 0.857307249712313, 'recall': 0.9118727050183598, 'f1': 0.8837485172004744, 'number': 817} | {'precision': 0.6761904761904762, 'recall': 0.5966386554621849, 'f1': 0.6339285714285715, 'number': 119} | {'precision': 0.8941068139963168, 'recall': 0.9015784586815228, 'f1': 0.8978270920018492, 'number': 1077} | 0.8675 | 0.8877 | 0.8775 | 0.8242 |
| 0.0014 | 84.21 | 1600 | 1.4741 | {'precision': 0.8871359223300971, 'recall': 0.8947368421052632, 'f1': 0.890920170627666, 'number': 817} | {'precision': 0.7590361445783133, 'recall': 0.5294117647058824, 'f1': 0.6237623762376238, 'number': 119} | {'precision': 0.8777969018932874, 'recall': 0.947075208913649, 'f1': 0.9111210361768646, 'number': 1077} | 0.8768 | 0.9011 | 0.8888 | 0.8407 |
| 0.0005 | 94.74 | 1800 | 1.5542 | {'precision': 0.871824480369515, 'recall': 0.9241126070991432, 'f1': 0.8972073677956032, 'number': 817} | {'precision': 0.7111111111111111, 'recall': 0.5378151260504201, 'f1': 0.6124401913875598, 'number': 119} | {'precision': 0.9029038112522686, 'recall': 0.9238625812441968, 'f1': 0.9132629646626893, 'number': 1077} | 0.8814 | 0.9011 | 0.8912 | 0.8219 |
| 0.0008 | 105.26 | 2000 | 1.4726 | {'precision': 0.8964677222898904, 'recall': 0.9008567931456548, 'f1': 0.8986568986568988, 'number': 817} | {'precision': 0.7446808510638298, 'recall': 0.5882352941176471, 'f1': 0.6572769953051643, 'number': 119} | {'precision': 0.8958517210944396, 'recall': 0.9424326833797586, 'f1': 0.918552036199095, 'number': 1077} | 0.8892 | 0.9046 | 0.8968 | 0.8387 |
| 0.0003 | 115.79 | 2200 | 1.5233 | {'precision': 0.8910179640718563, 'recall': 0.9106487148102815, 'f1': 0.900726392251816, 'number': 817} | {'precision': 0.71, 'recall': 0.5966386554621849, 'f1': 0.6484018264840181, 'number': 119} | {'precision': 0.9049773755656109, 'recall': 0.9285051067780873, 'f1': 0.916590284142988, 'number': 1077} | 0.8897 | 0.9016 | 0.8956 | 0.8354 |
| 0.0001 | 126.32 | 2400 | 1.5261 | {'precision': 0.8817966903073287, 'recall': 0.9130966952264382, 'f1': 0.8971737823211066, 'number': 817} | {'precision': 0.7319587628865979, 'recall': 0.5966386554621849, 'f1': 0.6574074074074073, 'number': 119} | {'precision': 0.8998194945848376, 'recall': 0.9257195914577531, 'f1': 0.9125858123569794, 'number': 1077} | 0.8844 | 0.9011 | 0.8927 | 0.8362 |
Framework versions
- Transformers 4.27.1
- Pytorch 1.13.1+cu116
- Datasets 2.10.1
- Tokenizers 0.13.2
- Downloads last month
- 5
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.