git-large-r-coco-IDB_ADv1_COCOv6-rv3

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: 0.0863
• Meteor Score: {'meteor': 0.6759040451749316}

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: 6
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 96
• 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_with_restarts
• lr_scheduler_warmup_steps: 50
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Meteor Score
0.5819	5.0	5	0.0931	{'meteor': 0.6828111433072758}
0.5633	10.0	10	0.0926	{'meteor': 0.67882651810593}
0.5255	15.0	15	0.0886	{'meteor': 0.6786692018375606}
0.4719	20.0	20	0.0863	{'meteor': 0.6759040451749316}

Framework versions

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