git-large-r-coco-IDB2-V2-IDB_ADv1

This model is a fine-tuned version of ooliverz/git-large-r-coco-IDB2-V2 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.0528
• Meteor Score: {'meteor': 0.48099544022939256}

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: 3e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 1024
• 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
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Meteor Score
0.0752	5.0	5	0.0707	{'meteor': 0.45989970933738533}
0.0562	10.0	10	0.0602	{'meteor': 0.45979149509400175}
0.0436	15.0	15	0.0567	{'meteor': 0.4674583504146179}
0.0346	20.0	20	0.0530	{'meteor': 0.4749086336659367}
0.028	25.0	25	0.0524	{'meteor': 0.4770807569254236}
0.0232	30.0	30	0.0512	{'meteor': 0.4786903333280637}
0.0196	35.0	35	0.0508	{'meteor': 0.48157768145979507}
0.0171	40.0	40	0.0512	{'meteor': 0.48242689217394297}
0.0152	45.0	45	0.0515	{'meteor': 0.48273124613648705}
0.0139	50.0	50	0.0517	{'meteor': 0.4813023884740672}
0.0129	55.0	55	0.0519	{'meteor': 0.48118479220464483}
0.0122	60.0	60	0.0521	{'meteor': 0.48099544022939256}
0.0115	65.0	65	0.0522	{'meteor': 0.4803039037483629}
0.0112	70.0	70	0.0525	{'meteor': 0.4806395037000483}
0.0106	75.0	75	0.0527	{'meteor': 0.48061573437081007}
0.0105	80.0	80	0.0528	{'meteor': 0.48061573437081007}
0.0105	85.0	85	0.0528	{'meteor': 0.4809716709001543}
0.0105	90.0	90	0.0528	{'meteor': 0.4809716709001543}
0.0103	95.0	95	0.0528	{'meteor': 0.48099544022939256}
0.0102	100.0	100	0.0528	{'meteor': 0.48099544022939256}

Framework versions

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