lilt-roberta-en-base-finetuned-funsd

nielsr/lilt-roberta-en-base-finetuned-funsd

lilt-roberta-en-base-finetuned-funsd is a token classification model from nielsr. Use it when you need labels on individual words, such as names. It is set up for transformers.

should probably proofread and complete it, then remove this comment. --

Downloads · 30 days

243

1% of all-time downloads

All-time downloads

24.2K

Public

Parameters

130M

1.6 GB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.bin521 MB · 50%

Parameter types

How the weights are stored.

F32130M · 100%

Try a prompt

Base models

Datasets

funsd-layoutlmv3

Model card

From the Hugging Face model README

lilt-roberta-en-base-finetuned-funsd

This model is a fine-tuned version of nielsr/lilt-roberta-en-base on the funsd-layoutlmv3 dataset. It achieves the following results on the evaluation set:

Loss: 1.6552
Precision: 0.8762
Recall: 0.8857
F1: 0.8809
Accuracy: 0.8068

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
lr_scheduler_warmup_steps: 0.1
training_steps: 2000

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	5.26	100	1.1789	0.8506	0.8485	0.8495	0.7869
No log	10.53	200	1.2382	0.8360	0.8788	0.8569	0.7970
No log	15.79	300	1.3766	0.8557	0.8897	0.8724	0.7909
No log	21.05	400	1.5590	0.8368	0.8763	0.8561	0.7792
0.04	26.32	500	1.4379	0.8562	0.8813	0.8685	0.7992
0.04	31.58	600	1.5397	0.8593	0.8947	0.8766	0.8054
0.04	36.84	700	1.6132	0.8621	0.8723	0.8672	0.7933
0.04	42.11	800	1.6483	0.8566	0.8872	0.8716	0.7777
0.04	47.37	900	1.6593	0.8641	0.8813	0.8726	0.7895
0.0044	52.63	1000	1.6704	0.8595	0.8718	0.8656	0.7925
0.0044	57.89	1100	1.6795	0.8495	0.8803	0.8646	0.7748
0.0044	63.16	1200	1.5515	0.8604	0.8912	0.8755	0.7991
0.0044	68.42	1300	1.6665	0.8573	0.8867	0.8718	0.7821
0.0044	73.68	1400	1.5893	0.8604	0.8877	0.8738	0.7895
0.0008	78.95	1500	1.5613	0.8603	0.8872	0.8736	0.8123
0.0008	84.21	1600	1.5853	0.8521	0.8872	0.8693	0.8040
0.0008	89.47	1700	1.6539	0.8707	0.8833	0.8769	0.8077
0.0008	94.74	1800	1.6634	0.8787	0.8813	0.8800	0.8079
0.0008	100.0	1900	1.6534	0.8810	0.8862	0.8836	0.8073
0.0004	105.26	2000	1.6552	0.8762	0.8857	0.8809	0.8068

Framework versions

Transformers 4.23.0.dev0
Pytorch 1.12.1+cu113
Datasets 2.5.1
Tokenizers 0.13.0

Task: Token Classification
Library: transformers
Type: lilt
Created: Sep 30, 2022
Updated: Dec 9, 2024