IC_ver6P_coco_swin_gpt2_50B_1e

This model is a fine-tuned version of VK246/IC_ver6O_coco_swin_gpt2_50A_1e on the coco dataset. It achieves the following results on the evaluation set:

• Loss: 0.9083
• Cider: 71.9821
• Rouge1: 40.8231
• Rouge2: 15.3858
• Rougel: 36.9192
• Rougelsum: 36.9129
• Bleu-1: 42.0506
• Bleu-2: 23.8307
• Bleu-3: 14.8432
• Bleu-4: 9.7337
• 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
0.3172	0.34	1000	1.0778	66.318	39.4397	14.0713	35.6057	35.6047	40.7003	22.4525	13.6705	8.7851	11.3063
0.4589	0.68	2000	0.9083	71.9821	40.8231	15.3858	36.9192	36.9129	42.0506	23.8307	14.8432	9.7337	11.3063

Framework versions

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