blip-image-captioning-base-finetuned-25epoch-1e-4

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

  • Loss: 1.5782
  • Wer Score: 0.3254
  • Blue Score: 0.6634

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: 0.0001
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 25

Training results

Training Loss Epoch Step Validation Loss Wer Score Blue Score
6.8732 1.01 35 3.2765 0.9105 0.0244
2.1287 2.03 70 1.7647 0.7923 0.1135
1.719 3.04 105 1.6799 0.6832 0.2382
1.6093 4.06 140 1.6328 0.5892 0.3339
1.5421 5.07 175 1.6058 0.5249 0.4004
1.4871 6.09 210 1.5880 0.4492 0.5001
1.4519 7.1 245 1.5773 0.4079 0.5568
1.4275 8.12 280 1.5694 0.3794 0.5908
1.4115 9.13 315 1.5696 0.3692 0.6075
1.3992 10.14 350 1.5649 0.3477 0.6326
1.3895 11.16 385 1.5662 0.3437 0.6424
1.3843 12.17 420 1.5694 0.3430 0.6402
1.3793 13.19 455 1.5692 0.3336 0.6553
1.3764 14.2 490 1.5710 0.3288 0.6571
1.3743 15.22 525 1.5711 0.3280 0.6631
1.373 16.23 560 1.5724 0.3269 0.6617
1.3715 17.25 595 1.5736 0.3241 0.6654
1.3704 18.26 630 1.5752 0.3271 0.6618
1.37 19.28 665 1.5758 0.3258 0.6629
1.3694 20.29 700 1.5768 0.3264 0.6627
1.3687 21.3 735 1.5767 0.3258 0.6640
1.3689 22.32 770 1.5775 0.3250 0.6636
1.3684 23.33 805 1.5780 0.3258 0.6636
1.3682 24.35 840 1.5782 0.3254 0.6634

Framework versions

  • Transformers 4.30.2
  • Pytorch 2.0.1+cu118
  • Datasets 2.14.3
  • Tokenizers 0.13.3
Downloads last month
15
Inference API
Inference API (serverless) does not yet support transformers models for this pipeline type.