IC_ver6L_coco_swin_gpt2_50B_1e

This model is a fine-tuned version of VK246/IC_ver6K_coco_swin_gpt2_50A_1e on the coco dataset. It achieves the following results on the evaluation set:

• Loss: 0.8371
• Cider: 72.6054
• Rouge1: 41.2906
• Rouge2: 15.8851
• Rougel: 37.3963
• Rougelsum: 37.4014
• Bleu-1: 42.3937
• Bleu-2: 24.3104
• Bleu-3: 15.291
• Bleu-4: 10.0894
• 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	Cider	Rouge1	Rouge2	Rougel	Rougelsum	Bleu-1	Bleu-2	Bleu-3	Bleu-4	Gen Len
0.4535	0.34	1000	0.9282	70.2558	40.6455	15.1383	36.6998	36.6999	41.838	23.6151	14.6823	9.6083	11.3063
0.5716	0.68	2000	0.8371	72.6054	41.2906	15.8851	37.3963	37.4014	42.3937	24.3104	15.291	10.0894	11.3063

Framework versions

• Transformers 4.32.0
• Pytorch 2.0.1+cu118
• Datasets 2.14.4
• Tokenizers 0.13.3