synergycodes/diagram_detr_r50_finetuned

daigram_detr_r50_albumentations_finetuning

This model is a fine-tuned version of kacper-cierzniewski/daigram_detr_r50_albumentations on the bpmn-shapes dataset. It achieves the following results on the evaluation set:

• Loss: 0.9817

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: 1e-05
• train_batch_size: 48
• eval_batch_size: 48
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.9457	12.5	50	1.0238
0.9717	25.0	100	1.0411
0.9823	37.5	150	1.0269
0.9524	50.0	200	1.0518
0.9886	62.5	250	1.0548
0.9638	75.0	300	1.0454
0.948	87.5	350	1.0240
0.9312	100.0	400	1.0281
0.9183	112.5	450	1.0112
0.9219	125.0	500	1.0110
0.9285	137.5	550	1.0325
0.9177	150.0	600	1.0009
0.9323	162.5	650	1.0124
0.9333	175.0	700	1.0154
0.9386	187.5	750	1.0188
0.9586	200.0	800	0.9978
0.894	212.5	850	1.0087
0.8999	225.0	900	1.0055
0.9324	237.5	950	1.0185
0.9313	250.0	1000	0.9840
0.9177	262.5	1050	0.9785
0.8918	275.0	1100	0.9874
0.9145	287.5	1150	0.9802
0.89	300.0	1200	0.9879
0.8818	312.5	1250	0.9857
0.9256	325.0	1300	0.9951
0.9028	337.5	1350	1.0001
0.9252	350.0	1400	1.0033
0.9017	362.5	1450	0.9916
0.8783	375.0	1500	0.9858
0.911	387.5	1550	0.9758
0.8797	400.0	1600	0.9810
0.8995	412.5	1650	0.9840
0.8781	425.0	1700	0.9843
0.8897	437.5	1750	0.9745
0.905	450.0	1800	0.9825
0.8961	462.5	1850	0.9781
0.8865	475.0	1900	0.9781
0.8824	487.5	1950	0.9794
0.8836	500.0	2000	0.9817

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0