Vit-GPT2-UCA-UCF-06

This model is a fine-tuned version of nlpconnect/vit-gpt2-image-captioning on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1937
• Rouge1: 29.6433
• Rouge2: 8.3589
• Rougel: 25.256
• Rougelsum: 25.5825
• Gen Len: 15.63

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
0.8073	0.3258	500	0.1840	31.5942	9.2754	27.0997	27.4879	17.309
0.6562	0.6516	1000	0.1805	31.3758	9.5474	26.788	27.1031	16.271
0.6123	0.9774	1500	0.1795	32.219	9.7783	27.4235	27.7537	16.455
0.5502	1.3030	2000	0.1821	31.0914	9.2688	26.5321	26.8962	15.66
0.5281	1.6288	2500	0.1832	31.0119	9.0876	26.4645	26.7925	16.042
0.5085	1.9546	3000	0.1847	31.0869	9.0206	26.2838	26.6729	16.004
0.4584	2.2802	3500	0.1919	29.6475	8.3551	25.1859	25.455	15.92
0.4536	2.6060	4000	0.1922	30.3476	8.7192	25.8444	26.0811	15.981
0.4477	2.9317	4500	0.1937	29.6433	8.3589	25.256	25.5825	15.63

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu121
• Datasets 3.3.1
• Tokenizers 0.21.0