blip-vqa-base-blip

This model is a fine-tuned version of Salesforce/blip-vqa-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4584
• Wer: 0.8802

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• 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: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.853	3.88	50	0.5155	0.8983
0.2416	7.72	100	0.3933	0.8862
0.0888	11.56	150	0.4088	0.8853
0.0402	15.4	200	0.4175	0.8793
0.0238	19.24	250	0.4288	0.8759
0.015	23.08	300	0.4322	0.875
0.0096	26.96	350	0.4346	0.8836
0.006	30.8	400	0.4446	0.8741
0.0039	34.64	450	0.4484	0.8767
0.0027	38.48	500	0.4536	0.8793
0.0019	42.32	550	0.4551	0.8793
0.0016	46.16	600	0.4573	0.8810
0.0014	50.0	650	0.4584	0.8802

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.2