IC_ver6c_coco_swin_gpt2_50Apc_1e

This model is a fine-tuned version of VK246/IC_ver6b_coco_swin_gpt2_50Bpc_1e on the coco dataset. It achieves the following results on the evaluation set:

• Loss: 0.7946
• Rouge1: 41.9094
• Rouge2: 16.3068
• Rougel: 38.073
• Rougelsum: 38.0746
• Bleu: 10.1966
• Gen Len: 11.2806

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
0.8232	0.17	500	0.8331	40.454	15.1311	36.7639	36.7714	9.2957	11.2806
0.8016	0.34	1000	0.8200	40.6374	15.5346	36.902	36.9055	9.6894	11.2806
0.8048	0.51	1500	0.8136	41.3382	15.9333	37.6502	37.6442	9.7743	11.2806
0.8018	0.68	2000	0.8028	41.5968	16.106	37.8326	37.836	9.9815	11.2806
0.8075	0.85	2500	0.7978	41.7017	16.1589	37.8899	37.8954	10.1244	11.2806

Framework versions

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