blip2-finetuned-vqa

This model is a fine-tuned version of ybelkada/blip2-opt-2.7b-fp16-sharded on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6786

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: 0.0002
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.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: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.7789	0.2051	1000	0.7354
0.6987	0.4102	2000	0.7221
0.6893	0.6153	3000	0.6991
0.6997	0.8204	4000	0.6786

Framework versions

• PEFT 0.15.2.dev0
• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1