git-large-r-coco-IDB2-V2-IDB2-V3

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: 0.5956

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: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• 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: 500
• num_epochs: 150
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.021	5.0	5	0.5467
0.029	10.0	10	0.5456
0.0285	15.0	15	0.5447
0.0262	20.0	20	0.5453
0.024	25.0	25	0.5478
0.0225	30.0	30	0.5512
0.0238	35.0	35	0.5544
0.0232	40.0	40	0.5563
0.0218	45.0	45	0.5566
0.0213	50.0	50	0.5565
0.0198	55.0	55	0.5571
0.0191	60.0	60	0.5592
0.0175	65.0	65	0.5620
0.0171	70.0	70	0.5629
0.0158	75.0	75	0.5648
0.0146	80.0	80	0.5667
0.0135	85.0	85	0.5683
0.0127	90.0	90	0.5711
0.012	95.0	95	0.5720
0.011	100.0	100	0.5738
0.0096	105.0	105	0.5758
0.0086	110.0	110	0.5781
0.008	115.0	115	0.5785
0.0073	120.0	120	0.5801
0.0069	125.0	125	0.5810
0.0059	130.0	130	0.5868
0.0052	135.0	135	0.5857
0.005	140.0	140	0.5875
0.0043	145.0	145	0.5960
0.0039	150.0	150	0.5956

Framework versions

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