image-captioning-output

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: 0.5164
• Rouge1: 35.5267
• Rouge2: 12.254
• Rougel: 32.968
• Rougelsum: 32.9723
• Gen Len: 12.395

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
0.5193	0.25	500	0.5171	33.0319	10.364	30.6939	30.6888	12.1
0.4842	0.5	1000	0.5102	33.7318	10.8199	31.1842	31.18	11.3
0.4724	0.75	1500	0.5028	34.6981	11.4074	31.9128	31.9158	12.02
0.4632	1.0	2000	0.5012	35.9443	12.8742	33.4061	33.377	11.04
0.377	1.25	2500	0.5026	35.7745	12.2309	33.3234	33.3353	11.735
0.3819	1.5	3000	0.5018	36.0145	13.0296	33.5985	33.6182	12.285
0.3788	1.75	3500	0.5030	35.9016	12.5276	33.4995	33.5033	11.305
0.3654	2.0	4000	0.5020	36.2476	12.945	33.6453	33.6595	11.9
0.3102	2.25	4500	0.5146	36.1507	13.0072	33.3889	33.3786	12.305
0.3137	2.5	5000	0.5166	35.7413	12.5693	33.2646	33.2508	12.71
0.3111	2.75	5500	0.5171	35.5658	12.511	33.0581	33.0518	12.55
0.3023	3.0	6000	0.5164	35.5267	12.254	32.968	32.9723	12.395

Framework versions

• Transformers 4.40.0
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1