IC_ver6b_coco_swin_gpt2_50Bpc_1e

This model is a fine-tuned version of VK246/IC_ver6a_coco_swin_gpt2_50Apc_1e on the coco dataset. It achieves the following results on the evaluation set:

• Loss: 0.8180
• Rouge1: 41.462
• Rouge2: 16.1291
• Rougel: 37.6518
• Rougelsum: 37.6471
• Bleu: 9.9643
• Gen Len: 11.3063

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: 96
• eval_batch_size: 96
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Bleu	Gen Len
0.8889	0.17	500	0.8659	39.7654	14.603	36.1709	36.1638	8.8491	11.3063
0.8756	0.34	1000	0.8515	40.3678	15.2852	36.7303	36.7188	9.3029	11.3063
0.862	0.51	1500	0.8388	40.7537	15.2635	37.0523	37.0379	9.3057	11.3063
0.8546	0.68	2000	0.8281	40.961	15.6192	37.1627	37.1546	9.7453	11.3063
0.837	0.85	2500	0.8214	41.5703	16.1006	37.7767	37.7654	9.9062	11.3063

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu118
• Datasets 2.13.1
• Tokenizers 0.13.3