bros-funsd-finetuned

This model is a fine-tuned version of naver-clova-ocr/bros-base-uncased on the funsd dataset. It achieves the following results on the evaluation set:

• Loss: 1.7866
• Precision: 0.5993
• Recall: 0.6416
• F1: 0.6197
• Accuracy: 0.7016

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	10	1.6503	0.0207	0.0032	0.0055	0.3213
No log	2.0	20	1.5622	0.1480	0.0596	0.0850	0.3890
No log	3.0	30	1.5357	0.0770	0.0672	0.0717	0.3803
No log	4.0	40	1.5160	0.1058	0.0976	0.1015	0.4078
No log	5.0	50	1.4925	0.1608	0.1768	0.1684	0.4354
No log	6.0	60	1.4216	0.2011	0.2288	0.2141	0.4571
No log	7.0	70	1.3546	0.2565	0.3241	0.2864	0.5001
No log	8.0	80	1.2950	0.2829	0.3818	0.3250	0.5048
No log	9.0	90	1.2862	0.2909	0.3745	0.3275	0.5226
No log	10.0	100	1.2108	0.2911	0.3815	0.3302	0.5491
No log	11.0	110	1.2023	0.3348	0.3609	0.3474	0.5545
No log	12.0	120	1.1720	0.3616	0.4030	0.3812	0.5668
No log	13.0	130	1.1267	0.3600	0.4005	0.3792	0.5825
No log	14.0	140	1.1025	0.3677	0.4499	0.4047	0.6144
No log	15.0	150	1.1038	0.3914	0.4655	0.4252	0.6182
No log	16.0	160	1.1034	0.4144	0.4769	0.4434	0.6399
No log	17.0	170	1.1885	0.4136	0.5250	0.4627	0.6303
No log	18.0	180	1.1734	0.4652	0.4854	0.4751	0.6491
No log	19.0	190	1.2263	0.4312	0.5995	0.5016	0.6457
No log	20.0	200	1.2326	0.4482	0.5612	0.4984	0.6478
No log	21.0	210	1.1374	0.4892	0.5954	0.5371	0.6776
No log	22.0	220	1.2278	0.4939	0.5779	0.5326	0.6712
No log	23.0	230	1.2979	0.4728	0.6030	0.5300	0.6642
No log	24.0	240	1.3170	0.4885	0.5916	0.5351	0.6682
No log	25.0	250	1.3692	0.4746	0.6011	0.5304	0.6596
No log	26.0	260	1.3706	0.5121	0.6106	0.5570	0.6742
No log	27.0	270	1.4494	0.5195	0.6036	0.5584	0.6719
No log	28.0	280	1.4790	0.5207	0.6027	0.5587	0.6678
No log	29.0	290	1.4106	0.5499	0.5887	0.5686	0.6838
No log	30.0	300	1.4539	0.5607	0.5954	0.5775	0.6810
No log	31.0	310	1.4746	0.5681	0.5989	0.5831	0.6827
No log	32.0	320	1.5373	0.5233	0.6144	0.5652	0.6698
No log	33.0	330	1.6007	0.5131	0.6353	0.5677	0.6682
No log	34.0	340	1.5237	0.5392	0.6489	0.5890	0.6868
No log	35.0	350	1.5382	0.5439	0.6239	0.5812	0.6908
No log	36.0	360	1.5363	0.5615	0.6071	0.5834	0.6872
No log	37.0	370	1.5504	0.5572	0.6201	0.5870	0.6943
No log	38.0	380	1.6496	0.5478	0.6176	0.5806	0.6796
No log	39.0	390	1.6083	0.5665	0.6144	0.5895	0.6913
No log	40.0	400	1.5588	0.5719	0.6239	0.5968	0.6977
No log	41.0	410	1.6280	0.5578	0.6328	0.5929	0.6928
No log	42.0	420	1.5925	0.5842	0.6112	0.5974	0.7023
No log	43.0	430	1.5921	0.5810	0.6204	0.6001	0.6981
No log	44.0	440	1.6152	0.5740	0.6207	0.5964	0.6917
No log	45.0	450	1.6629	0.5634	0.6283	0.5941	0.6853
No log	46.0	460	1.6112	0.5829	0.6214	0.6015	0.7021
No log	47.0	470	1.6214	0.5761	0.6258	0.5999	0.6982
No log	48.0	480	1.6216	0.5953	0.6119	0.6034	0.7023
No log	49.0	490	1.6592	0.5809	0.6163	0.5981	0.6962
0.4349	50.0	500	1.6796	0.5603	0.6489	0.6014	0.6947
0.4349	51.0	510	1.6835	0.5967	0.6001	0.5984	0.6933
0.4349	52.0	520	1.6615	0.5832	0.6553	0.6171	0.6999
0.4349	53.0	530	1.6553	0.5778	0.6565	0.6147	0.6970
0.4349	54.0	540	1.6980	0.5946	0.6004	0.5975	0.6888
0.4349	55.0	550	1.6484	0.5694	0.6356	0.6007	0.6960
0.4349	56.0	560	1.6996	0.5902	0.6293	0.6091	0.6941
0.4349	57.0	570	1.6973	0.5780	0.6337	0.6046	0.6947
0.4349	58.0	580	1.7212	0.5973	0.6087	0.6030	0.6969
0.4349	59.0	590	1.7086	0.5791	0.6435	0.6096	0.6976
0.4349	60.0	600	1.6767	0.5845	0.6233	0.6033	0.6996
0.4349	61.0	610	1.6744	0.5886	0.6201	0.6039	0.6993
0.4349	62.0	620	1.6783	0.5989	0.6286	0.6134	0.6999
0.4349	63.0	630	1.6958	0.5936	0.6489	0.6200	0.7019
0.4349	64.0	640	1.7297	0.5806	0.6286	0.6037	0.6941
0.4349	65.0	650	1.7373	0.5804	0.6540	0.6150	0.6961
0.4349	66.0	660	1.7579	0.5818	0.6404	0.6097	0.6941
0.4349	67.0	670	1.7654	0.5889	0.6369	0.6120	0.6971
0.4349	68.0	680	1.7649	0.5846	0.6515	0.6162	0.6953
0.4349	69.0	690	1.7294	0.5940	0.6445	0.6182	0.6999
0.4349	70.0	700	1.7256	0.5871	0.6511	0.6175	0.7021
0.4349	71.0	710	1.7303	0.5889	0.6518	0.6187	0.7029
0.4349	72.0	720	1.7391	0.5994	0.6334	0.6159	0.7023
0.4349	73.0	730	1.7270	0.5838	0.6448	0.6128	0.6999
0.4349	74.0	740	1.7357	0.6060	0.6324	0.6189	0.7035
0.4349	75.0	750	1.7210	0.6030	0.6362	0.6192	0.7036
0.4349	76.0	760	1.7575	0.5903	0.6473	0.6175	0.6990
0.4349	77.0	770	1.7530	0.5859	0.6416	0.6125	0.6958
0.4349	78.0	780	1.7395	0.5865	0.6445	0.6141	0.6988
0.4349	79.0	790	1.7432	0.5900	0.6575	0.6219	0.7025
0.4349	80.0	800	1.7497	0.5957	0.6556	0.6242	0.7039
0.4349	81.0	810	1.7590	0.6003	0.6467	0.6226	0.7040
0.4349	82.0	820	1.7641	0.5979	0.6413	0.6189	0.7019
0.4349	83.0	830	1.7632	0.6103	0.6407	0.6251	0.7070
0.4349	84.0	840	1.7602	0.6082	0.6420	0.6246	0.7066
0.4349	85.0	850	1.7697	0.6014	0.6458	0.6228	0.7051
0.4349	86.0	860	1.7828	0.5945	0.6397	0.6163	0.7001
0.4349	87.0	870	1.7834	0.6005	0.6369	0.6182	0.7005
0.4349	88.0	880	1.7760	0.5966	0.6388	0.6170	0.7013
0.4349	89.0	890	1.7757	0.5942	0.6426	0.6174	0.7021
0.4349	90.0	900	1.7755	0.5946	0.6442	0.6184	0.7025
0.4349	91.0	910	1.7778	0.5964	0.6432	0.6189	0.7012
0.4349	92.0	920	1.7757	0.5993	0.6435	0.6206	0.7019
0.4349	93.0	930	1.7751	0.6014	0.6448	0.6223	0.7025
0.4349	94.0	940	1.7769	0.6024	0.6410	0.6211	0.7025
0.4349	95.0	950	1.7791	0.6026	0.6394	0.6204	0.7020
0.4349	96.0	960	1.7862	0.6016	0.6381	0.6193	0.7012
0.4349	97.0	970	1.7876	0.5985	0.6410	0.6190	0.7007
0.4349	98.0	980	1.7882	0.5976	0.6404	0.6182	0.7012
0.4349	99.0	990	1.7870	0.5988	0.6413	0.6193	0.7014
0.0052	100.0	1000	1.7866	0.5993	0.6416	0.6197	0.7016

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1