README.md

metadata

license: mit
tags:
  - generated_from_trainer
datasets:
  - funsd-layoutlmv3
model-index:
  - name: lilt-en-funsd
    results: []

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: 1.8784
• Answer: {'precision': 0.8651817116060961, 'recall': 0.9033047735618115, 'f1': 0.8838323353293414, 'number': 817}
• Header: {'precision': 0.6504854368932039, 'recall': 0.5630252100840336, 'f1': 0.6036036036036037, 'number': 119}
• Question: {'precision': 0.9073394495412844, 'recall': 0.9182915506035283, 'f1': 0.912782648823258, 'number': 1077}
• Overall Precision: 0.8768
• Overall Recall: 0.8912
• Overall F1: 0.8840
• Overall Accuracy: 0.7948

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

Training results

Training Loss	Epoch	Step	Validation Loss	Answer	Header	Question	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
0.4369	10.53	200	0.9022	{'precision': 0.8049065420560748, 'recall': 0.8433292533659731, 'f1': 0.8236700537955769, 'number': 817}	{'precision': 0.5317460317460317, 'recall': 0.5630252100840336, 'f1': 0.5469387755102041, 'number': 119}	{'precision': 0.8837420526793823, 'recall': 0.903435468895079, 'f1': 0.8934802571166208, 'number': 1077}	0.8301	0.8589	0.8442	0.7888
0.047	21.05	400	1.3222	{'precision': 0.8382526564344747, 'recall': 0.8690330477356181, 'f1': 0.8533653846153846, 'number': 817}	{'precision': 0.5447761194029851, 'recall': 0.6134453781512605, 'f1': 0.5770750988142292, 'number': 119}	{'precision': 0.8667866786678667, 'recall': 0.8941504178272981, 'f1': 0.8802559414990858, 'number': 1077}	0.8346	0.8674	0.8507	0.7837
0.015	31.58	600	1.4745	{'precision': 0.8549528301886793, 'recall': 0.8873929008567931, 'f1': 0.8708708708708709, 'number': 817}	{'precision': 0.5867768595041323, 'recall': 0.5966386554621849, 'f1': 0.5916666666666667, 'number': 119}	{'precision': 0.8755635707844905, 'recall': 0.9015784586815228, 'f1': 0.888380603842635, 'number': 1077}	0.8503	0.8778	0.8638	0.7969
0.0051	42.11	800	1.5719	{'precision': 0.8768472906403941, 'recall': 0.8714810281517748, 'f1': 0.8741559238796808, 'number': 817}	{'precision': 0.5736434108527132, 'recall': 0.6218487394957983, 'f1': 0.596774193548387, 'number': 119}	{'precision': 0.8794326241134752, 'recall': 0.9210770659238626, 'f1': 0.8997732426303855, 'number': 1077}	0.8594	0.8833	0.8711	0.7923
0.0041	52.63	1000	1.6771	{'precision': 0.8352402745995423, 'recall': 0.8935128518971848, 'f1': 0.8633944411590775, 'number': 817}	{'precision': 0.6568627450980392, 'recall': 0.5630252100840336, 'f1': 0.6063348416289592, 'number': 119}	{'precision': 0.8865116279069768, 'recall': 0.8848653667595172, 'f1': 0.8856877323420075, 'number': 1077}	0.8532	0.8693	0.8612	0.7877
0.0039	63.16	1200	1.6064	{'precision': 0.8609112709832134, 'recall': 0.8788249694002448, 'f1': 0.8697758933979407, 'number': 817}	{'precision': 0.6106194690265486, 'recall': 0.5798319327731093, 'f1': 0.5948275862068966, 'number': 119}	{'precision': 0.8897777777777778, 'recall': 0.9294336118848654, 'f1': 0.9091734786557675, 'number': 1077}	0.8629	0.8882	0.8754	0.8009
0.0019	73.68	1400	1.7674	{'precision': 0.8533178114086146, 'recall': 0.8971848225214198, 'f1': 0.8747016706443913, 'number': 817}	{'precision': 0.5769230769230769, 'recall': 0.5042016806722689, 'f1': 0.5381165919282511, 'number': 119}	{'precision': 0.8842676311030742, 'recall': 0.9080779944289693, 'f1': 0.8960146587265231, 'number': 1077}	0.8560	0.8798	0.8677	0.7981
0.0007	84.21	1600	1.8380	{'precision': 0.8469387755102041, 'recall': 0.9143206854345165, 'f1': 0.8793407886992348, 'number': 817}	{'precision': 0.6017699115044248, 'recall': 0.5714285714285714, 'f1': 0.5862068965517241, 'number': 119}	{'precision': 0.8931159420289855, 'recall': 0.9155060352831941, 'f1': 0.9041723979825768, 'number': 1077}	0.8580	0.8947	0.8760	0.7931
0.0007	94.74	1800	1.8108	{'precision': 0.8600478468899522, 'recall': 0.8800489596083231, 'f1': 0.8699334543254689, 'number': 817}	{'precision': 0.6435643564356436, 'recall': 0.5462184873949579, 'f1': 0.5909090909090908, 'number': 119}	{'precision': 0.8722849695916595, 'recall': 0.9322191272051996, 'f1': 0.9012567324955117, 'number': 1077}	0.8563	0.8882	0.8720	0.7887
0.0004	105.26	2000	1.9035	{'precision': 0.8627906976744186, 'recall': 0.9082007343941249, 'f1': 0.8849135360763267, 'number': 817}	{'precision': 0.6285714285714286, 'recall': 0.5546218487394958, 'f1': 0.5892857142857143, 'number': 119}	{'precision': 0.8955495004541326, 'recall': 0.9155060352831941, 'f1': 0.9054178145087237, 'number': 1077}	0.8683	0.8912	0.8796	0.7965
0.0002	115.79	2200	1.8784	{'precision': 0.8651817116060961, 'recall': 0.9033047735618115, 'f1': 0.8838323353293414, 'number': 817}	{'precision': 0.6504854368932039, 'recall': 0.5630252100840336, 'f1': 0.6036036036036037, 'number': 119}	{'precision': 0.9073394495412844, 'recall': 0.9182915506035283, 'f1': 0.912782648823258, 'number': 1077}	0.8768	0.8912	0.8840	0.7948
0.0002	126.32	2400	1.9075	{'precision': 0.8640093786635404, 'recall': 0.9020807833537332, 'f1': 0.8826347305389222, 'number': 817}	{'precision': 0.6296296296296297, 'recall': 0.5714285714285714, 'f1': 0.5991189427312775, 'number': 119}	{'precision': 0.9041970802919708, 'recall': 0.9201485608170845, 'f1': 0.9121030832949838, 'number': 1077}	0.8731	0.8922	0.8826	0.7959

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu118
• Datasets 2.13.0
• Tokenizers 0.13.3