blip2_lora_vqa_model

This model is a fine-tuned version of Salesforce/blip2-flan-t5-xl on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0599
• Exact: 75.7776
• F1: 79.5668
• Total: 1061
• Hasans Exact: 75.7776
• Hasans F1: 79.5668
• Hasans Total: 1061
• Best Exact: 75.7776
• Best Exact Thresh: 0.0
• Best F1: 79.5668
• Best F1 Thresh: 0.0

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: 0.002
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Exact	F1	Total	Hasans Exact	Hasans F1	Hasans Total	Best Exact	Best Exact Thresh	Best F1	Best F1 Thresh
No log	1.0	77	0.1726	57.6814	62.4231	1061	57.6814	62.4231	1061	57.6814	0.0	62.4231	0.0
1.8252	2.0	154	0.1175	64.3732	69.3316	1061	64.3732	69.3316	1061	64.3732	0.0	69.3316	0.0
0.1488	3.0	231	0.0969	65.6927	69.8246	1061	65.6927	69.8246	1061	65.6927	0.0	69.8246	0.0
0.1133	4.0	308	0.0835	69.3685	74.0934	1061	69.3685	74.0934	1061	69.3685	0.0	74.0934	0.0
0.1133	5.0	385	0.0741	71.3478	75.5583	1061	71.3478	75.5583	1061	71.3478	0.0	75.5583	0.0
0.0912	6.0	462	0.0661	71.7248	76.0127	1061	71.7248	76.0127	1061	71.7248	0.0	76.0127	0.0
0.0834	7.0	539	0.0691	72.7615	77.0162	1061	72.7615	77.0162	1061	72.7615	0.0	77.0162	0.0
0.0686	8.0	616	0.0632	74.0811	77.5827	1061	74.0811	77.5827	1061	74.0811	0.0	77.5827	0.0
0.0686	9.0	693	0.0609	74.8351	78.6302	1061	74.8351	78.6302	1061	74.8351	0.0	78.6302	0.0
0.0626	10.0	770	0.0599	75.7776	79.5668	1061	75.7776	79.5668	1061	75.7776	0.0	79.5668	0.0

Framework versions

• PEFT 0.13.2
• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3