README.md

metadata

library_name: transformers
license: apache-2.0
base_model: google/vit-base-patch16-384
tags:
  - generated_from_trainer
datasets:
  - webdataset
metrics:
  - accuracy
  - f1
  - precision
  - recall
model-index:
  - name: frost-vision-v2-google_vit-base-patch16-384-v2024-11-10
    results:
      - task:
          name: Image Classification
          type: image-classification
        dataset:
          name: webdataset
          type: webdataset
          config: default
          split: train
          args: default
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.9274647887323944
          - name: F1
            type: f1
            value: 0.8186619718309859
          - name: Precision
            type: precision
            value: 0.8172231985940246
          - name: Recall
            type: recall
            value: 0.8201058201058201

frost-vision-v2-google_vit-base-patch16-384-v2024-11-10

This model is a fine-tuned version of google/vit-base-patch16-384 on the webdataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.1847
• Accuracy: 0.9275
• F1: 0.8187
• Precision: 0.8172
• Recall: 0.8201

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.2243	1.4085	100	0.2088	0.9243	0.7981	0.8534	0.7496
0.2438	2.8169	200	0.1819	0.9299	0.8103	0.8817	0.7496
0.1338	4.2254	300	0.1608	0.9377	0.8449	0.8397	0.8501
0.1224	5.6338	400	0.1735	0.9271	0.8179	0.8158	0.8201
0.1065	7.0423	500	0.1847	0.9275	0.8187	0.8172	0.8201
0.1008	8.4507	600	0.1710	0.9405	0.8506	0.8528	0.8483
0.1005	9.8592	700	0.1823	0.9384	0.8405	0.8698	0.8131
0.0756	11.2676	800	0.1771	0.9415	0.8520	0.8613	0.8430
0.0653	12.6761	900	0.1971	0.9324	0.8310	0.8295	0.8325
0.0367	14.0845	1000	0.2123	0.9296	0.8221	0.8294	0.8148
0.0459	15.4930	1100	0.2006	0.9335	0.832	0.8387	0.8254
0.0559	16.9014	1200	0.2097	0.9313	0.8232	0.8470	0.8007
0.0382	18.3099	1300	0.2055	0.9352	0.8372	0.8401	0.8342
0.0361	19.7183	1400	0.2070	0.9335	0.8305	0.8449	0.8166
0.0358	21.1268	1500	0.1959	0.9398	0.8458	0.8653	0.8272
0.0382	22.5352	1600	0.2097	0.9320	0.8269	0.8412	0.8131
0.0285	23.9437	1700	0.2016	0.9415	0.8515	0.8639	0.8395
0.0141	25.3521	1800	0.2161	0.9366	0.8384	0.8537	0.8236
0.0179	26.7606	1900	0.2073	0.9377	0.8427	0.8495	0.8360
0.0263	28.1690	2000	0.2097	0.9391	0.8457	0.8556	0.8360
0.0191	29.5775	2100	0.2101	0.9377	0.8415	0.8545	0.8289

Framework versions

• Transformers 4.44.2
• Pytorch 2.5.0+cu121
• Datasets 3.1.0
• Tokenizers 0.19.1