git-large-r-coco-IDB_ADv1_COCOv5-r

This model is a fine-tuned version of microsoft/git-large-r-coco on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.1023
• Meteor Score: {'meteor': 0.6267975522399738}

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: 128
• eval_batch_size: 128
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 256
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 50
• num_epochs: 230
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Meteor Score
11.5017	5.0	5	11.5369	{'meteor': 0.045368587323702074}
10.8572	10.0	10	10.0742	{'meteor': 0.04593520140957956}
9.8764	15.0	15	9.4628	{'meteor': 0.05771587153621692}
9.2502	20.0	20	8.8179	{'meteor': 0.06110521825172792}
8.6276	25.0	25	8.1813	{'meteor': 0.07246240749923287}
8.0194	30.0	30	7.6100	{'meteor': 0.07533363641176034}
7.477	35.0	35	7.1132	{'meteor': 0.10920563586662041}
6.9816	40.0	40	6.6324	{'meteor': 0.14374157554815933}
6.4978	45.0	45	6.2079	{'meteor': 0.15418376317939878}
6.0276	50.0	50	5.6487	{'meteor': 0.1699020309800082}
5.4772	55.0	55	5.0680	{'meteor': 0.36037584656137706}
4.9039	60.0	60	4.4911	{'meteor': 0.40338651921680374}
4.335	65.0	65	3.9227	{'meteor': 0.43118679398975585}
3.7748	70.0	70	3.3636	{'meteor': 0.4463490343981121}
3.2244	75.0	75	2.8195	{'meteor': 0.45737949916708276}
2.6899	80.0	80	2.2988	{'meteor': 0.4669755898073271}
2.1812	85.0	85	1.8153	{'meteor': 0.4724530704117843}
1.7092	90.0	90	1.3812	{'meteor': 0.48263991671863704}
1.2887	95.0	95	1.0136	{'meteor': 0.49822448249218215}
0.9349	100.0	100	0.7227	{'meteor': 0.509467191188813}
0.6561	105.0	105	0.5085	{'meteor': 0.5351147967011172}
0.4506	110.0	110	0.3610	{'meteor': 0.5574306081501034}
0.3075	115.0	115	0.2642	{'meteor': 0.5675947880800101}
0.213	120.0	120	0.2022	{'meteor': 0.5850311864966891}
0.1489	125.0	125	0.1618	{'meteor': 0.5678028139248027}
0.1143	130.0	130	0.1377	{'meteor': 0.5684491667111089}
0.0812	135.0	135	0.1182	{'meteor': 0.5897239512329581}
0.0583	140.0	140	0.1047	{'meteor': 0.5848418520612428}
0.0426	145.0	145	0.1006	{'meteor': 0.5917122198152872}
0.0317	150.0	150	0.0969	{'meteor': 0.5924453353255843}
0.0265	155.0	155	0.0969	{'meteor': 0.6336499608699626}
0.0206	160.0	160	0.0959	{'meteor': 0.6173649413721484}
0.0168	165.0	165	0.0967	{'meteor': 0.616519373529578}
0.0136	170.0	170	0.0967	{'meteor': 0.6242033250759903}
0.0118	175.0	175	0.1001	{'meteor': 0.6214092096360933}
0.0105	180.0	180	0.1002	{'meteor': 0.627438846043814}
0.0094	185.0	185	0.1011	{'meteor': 0.6227117871180985}
0.0088	190.0	190	0.1012	{'meteor': 0.623899819923896}
0.0083	195.0	195	0.1020	{'meteor': 0.623696945492686}
0.0078	200.0	200	0.1023	{'meteor': 0.6248544615426194}
0.0076	205.0	205	0.1022	{'meteor': 0.6264504292056374}
0.0074	210.0	210	0.1022	{'meteor': 0.6271562698062381}
0.0073	215.0	215	0.1022	{'meteor': 0.6265034453166831}
0.0073	220.0	220	0.1022	{'meteor': 0.627146982870841}
0.0072	225.0	225	0.1023	{'meteor': 0.6263542605757835}
0.0071	230.0	230	0.1023	{'meteor': 0.6267975522399738}

Framework versions

• Transformers 4.46.1
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.20.2