git-large-r-coco-IDB_ADv1_COCOv4

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: 0.0985
• Meteor Score: {'meteor': 0.6045977816224402}

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: 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
• num_epochs: 200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Meteor Score
3.0153	5.0	5	2.9012	{'meteor': 0.11360374562500358}
2.6775	10.0	10	1.9996	{'meteor': 0.3522292411621663}
1.9479	15.0	15	1.6433	{'meteor': 0.4724950837310155}
1.6034	20.0	20	1.3467	{'meteor': 0.5315599885135005}
1.3119	25.0	25	1.0907	{'meteor': 0.5441775609077182}
1.0551	30.0	30	0.8677	{'meteor': 0.5627206125493585}
0.8324	35.0	35	0.6841	{'meteor': 0.5644140822761644}
0.6466	40.0	40	0.5342	{'meteor': 0.5714878241998325}
0.4953	45.0	45	0.4171	{'meteor': 0.5723224079693889}
0.3758	50.0	50	0.3280	{'meteor': 0.575759183770385}
0.2832	55.0	55	0.2617	{'meteor': 0.582651143305794}
0.2147	60.0	60	0.2141	{'meteor': 0.5823617638426744}
0.1637	65.0	65	0.1841	{'meteor': 0.587684204036701}
0.1283	70.0	70	0.1510	{'meteor': 0.5995580706928448}
0.097	75.0	75	0.1344	{'meteor': 0.5935823756573874}
0.0745	80.0	80	0.1212	{'meteor': 0.5971126553853467}
0.0576	85.0	85	0.1175	{'meteor': 0.6063465199710643}
0.0482	90.0	90	0.1095	{'meteor': 0.6256289206454945}
0.041	95.0	95	0.1021	{'meteor': 0.611542182926066}
0.0332	100.0	100	0.0967	{'meteor': 0.6332339587712169}
0.0263	105.0	105	0.0974	{'meteor': 0.6217214977486667}
0.021	110.0	110	0.0969	{'meteor': 0.6205827762386601}
0.0175	115.0	115	0.0974	{'meteor': 0.6149467166555415}
0.015	120.0	120	0.0969	{'meteor': 0.6141265010864785}
0.0132	125.0	125	0.0968	{'meteor': 0.6148825585221068}
0.0119	130.0	130	0.0947	{'meteor': 0.6099154157322018}
0.0109	135.0	135	0.0962	{'meteor': 0.6091804819899924}
0.0101	140.0	140	0.0967	{'meteor': 0.6079888261097917}
0.0096	145.0	145	0.0983	{'meteor': 0.6084492863562021}
0.0091	150.0	150	0.0982	{'meteor': 0.6061886051922543}
0.0088	155.0	155	0.0985	{'meteor': 0.6057257284449861}
0.0085	160.0	160	0.0988	{'meteor': 0.603016234486876}
0.0083	165.0	165	0.0984	{'meteor': 0.6042894593740327}
0.0082	170.0	170	0.0984	{'meteor': 0.6043600622898042}
0.0081	175.0	175	0.0985	{'meteor': 0.6043742843963279}
0.008	180.0	180	0.0985	{'meteor': 0.6041252208578323}
0.008	185.0	185	0.0986	{'meteor': 0.6043322953582039}
0.0079	190.0	190	0.0986	{'meteor': 0.6047551630278664}
0.0079	195.0	195	0.0985	{'meteor': 0.6048157030511828}
0.0079	200.0	200	0.0985	{'meteor': 0.6045977816224402}

Framework versions

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