git-large-r-coco-IDB_ADv1_COCOv3

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: 0.0966
• Meteor Score: {'meteor': 0.5408938263795505}

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: 32
• total_train_batch_size: 4096
• 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
3.2229	5.0	5	2.9789	{'meteor': 0.41552282487414405}
2.9387	10.0	10	2.6142	{'meteor': 0.4685053939421483}
2.592	15.0	15	2.2978	{'meteor': 0.4849322756130674}
2.2751	20.0	20	2.0019	{'meteor': 0.47420638870259907}
1.9738	25.0	25	1.7167	{'meteor': 0.48640146159813924}
1.6877	30.0	30	1.4518	{'meteor': 0.4898083781551464}
1.4214	35.0	35	1.2094	{'meteor': 0.487939608237839}
1.1769	40.0	40	0.9923	{'meteor': 0.4980774326833536}
0.9592	45.0	45	0.8040	{'meteor': 0.488335711225997}
0.7699	50.0	50	0.6462	{'meteor': 0.4919465483035675}
0.6114	55.0	55	0.5181	{'meteor': 0.4887190822781368}
0.481	60.0	60	0.4133	{'meteor': 0.4908850956442714}
0.3752	65.0	65	0.3325	{'meteor': 0.49324244604306894}
0.2921	70.0	70	0.2705	{'meteor': 0.49568446461184845}
0.2272	75.0	75	0.2250	{'meteor': 0.49361056097084616}
0.1776	80.0	80	0.1931	{'meteor': 0.5018863491081054}
0.1404	85.0	85	0.1653	{'meteor': 0.506990680377092}
0.1118	90.0	90	0.1468	{'meteor': 0.5071567168395572}
0.0913	95.0	95	0.1296	{'meteor': 0.5019147249270115}
0.0738	100.0	100	0.1212	{'meteor': 0.5151563978937971}
0.06	105.0	105	0.1144	{'meteor': 0.5254092884094786}
0.0494	110.0	110	0.1100	{'meteor': 0.5258164560531304}
0.0413	115.0	115	0.1046	{'meteor': 0.5305386302193077}
0.035	120.0	120	0.1005	{'meteor': 0.5405955416808903}
0.0301	125.0	125	0.1009	{'meteor': 0.5326314473682981}
0.0262	130.0	130	0.1016	{'meteor': 0.5466195008617952}
0.0233	135.0	135	0.0991	{'meteor': 0.5463553955219586}
0.0211	140.0	140	0.0987	{'meteor': 0.543975966495889}
0.0193	145.0	145	0.0979	{'meteor': 0.5269418821882673}
0.0179	150.0	150	0.0990	{'meteor': 0.5384964361269092}
0.0169	155.0	155	0.0985	{'meteor': 0.5374634680447925}
0.0161	160.0	160	0.0962	{'meteor': 0.5296357754808179}
0.0155	165.0	165	0.0968	{'meteor': 0.5335376085825252}
0.0151	170.0	170	0.0965	{'meteor': 0.5426897117754873}
0.0148	175.0	175	0.0971	{'meteor': 0.5435169793620099}
0.0146	180.0	180	0.0972	{'meteor': 0.5432382041598678}
0.0145	185.0	185	0.0969	{'meteor': 0.5413175577245792}
0.0144	190.0	190	0.0967	{'meteor': 0.5412533872506878}
0.0143	195.0	195	0.0966	{'meteor': 0.5407676909570709}
0.0143	200.0	200	0.0966	{'meteor': 0.5408938263795505}

Framework versions

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