README.md

metadata

library_name: transformers
license: apache-2.0
base_model: google/vit-base-patch16-224-in21k
tags:
  - generated_from_trainer
datasets:
  - webdataset
metrics:
  - accuracy
  - f1
  - precision
  - recall
model-index:
  - name: vit-base-patch16-224-in21k-v2025-1-31
    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: 0.8972972972972973
          - name: F1
            type: f1
            value: 0.7667958656330749
          - name: Precision
            type: precision
            value: 0.7866136514247847
          - name: Recall
            type: recall
            value: 0.7479521109010712

vit-base-patch16-224-in21k-v2025-1-31

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the webdataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.3391
• Accuracy: 0.8973
• F1: 0.7668
• Precision: 0.7866
• Recall: 0.7480

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: 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
• 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
0.4871	0.5682	100	0.4866	0.7903	0.1400	0.9449	0.0756
0.4151	1.1364	200	0.4007	0.8361	0.4540	0.9159	0.3018
0.3517	1.7045	300	0.3460	0.8671	0.6481	0.8060	0.5419
0.3337	2.2727	400	0.3202	0.8777	0.7034	0.7768	0.6427
0.3128	2.8409	500	0.2995	0.8774	0.6943	0.7940	0.6169
0.3199	3.4091	600	0.2980	0.8771	0.6960	0.7880	0.6232
0.3094	3.9773	700	0.3051	0.8764	0.7031	0.7679	0.6484
0.3068	4.5455	800	0.2753	0.8900	0.7409	0.7915	0.6963
0.3003	5.1136	900	0.2699	0.8890	0.7351	0.7973	0.6818
0.3012	5.6818	1000	0.2860	0.8799	0.7256	0.7495	0.7032
0.267	6.25	1100	0.2848	0.8832	0.7216	0.7812	0.6704
0.2364	6.8182	1200	0.2608	0.8896	0.7399	0.7903	0.6957
0.2401	7.3864	1300	0.2695	0.8885	0.7406	0.7798	0.7051
0.219	7.9545	1400	0.2599	0.8909	0.7413	0.7975	0.6925
0.1985	8.5227	1500	0.2668	0.8898	0.7421	0.7863	0.7026
0.1986	9.0909	1600	0.2762	0.8851	0.7316	0.7737	0.6938
0.1988	9.6591	1700	0.2765	0.8862	0.7404	0.7632	0.7190
0.167	10.2273	1800	0.2630	0.8940	0.7594	0.7788	0.7410
0.207	10.7955	1900	0.2637	0.8923	0.7557	0.7745	0.7379
0.1811	11.3636	2000	0.2568	0.8946	0.7609	0.7798	0.7429
0.171	11.9318	2100	0.2607	0.8935	0.7527	0.7906	0.7183
0.1571	12.5	2200	0.2552	0.8972	0.7708	0.7755	0.7662
0.1234	13.0682	2300	0.2676	0.8993	0.7694	0.7964	0.7442
0.1299	13.6364	2400	0.2683	0.8970	0.7655	0.7875	0.7448
0.1335	14.2045	2500	0.2823	0.8949	0.7559	0.7944	0.7209
0.1235	14.7727	2600	0.2753	0.8976	0.7671	0.7880	0.7473
0.1163	15.3409	2700	0.2884	0.8962	0.7644	0.7836	0.7461
0.1111	15.9091	2800	0.2770	0.8973	0.7675	0.7847	0.7511
0.1128	16.4773	2900	0.2773	0.8987	0.7722	0.7843	0.7606
0.0982	17.0455	3000	0.2754	0.8993	0.7716	0.7905	0.7536
0.1115	17.6136	3100	0.2956	0.8972	0.7640	0.7927	0.7372
0.07	18.1818	3200	0.2961	0.8977	0.7683	0.7863	0.7511
0.0993	18.75	3300	0.3041	0.8959	0.7639	0.7826	0.7461
0.0779	19.3182	3400	0.3012	0.9	0.7745	0.7889	0.7606
0.0691	19.8864	3500	0.3075	0.8964	0.7674	0.7784	0.7568
0.063	20.4545	3600	0.3271	0.8912	0.7509	0.7770	0.7265
0.0668	21.0227	3700	0.3229	0.8952	0.7649	0.7745	0.7555
0.0573	21.5909	3800	0.3236	0.8960	0.7626	0.7869	0.7398
0.0668	22.1591	3900	0.3251	0.8972	0.7629	0.7955	0.7328
0.062	22.7273	4000	0.3221	0.8987	0.7702	0.7895	0.7517
0.0647	23.2955	4100	0.3179	0.8959	0.7663	0.7767	0.7561
0.0417	23.8636	4200	0.3323	0.8969	0.7662	0.7847	0.7486
0.0623	24.4318	4300	0.3396	0.8945	0.7602	0.7804	0.7410
0.0361	25.0	4400	0.3418	0.8959	0.7623	0.7863	0.7398
0.0334	25.5682	4500	0.3404	0.8984	0.7703	0.7870	0.7543
0.0326	26.1364	4600	0.3376	0.8967	0.7676	0.7801	0.7555
0.052	26.7045	4700	0.3395	0.8972	0.7679	0.7827	0.7536
0.0341	27.2727	4800	0.3440	0.8953	0.7638	0.7783	0.7498
0.0459	27.8409	4900	0.3406	0.8980	0.7689	0.7869	0.7517
0.0392	28.4091	5000	0.3389	0.8977	0.7680	0.7870	0.7498
0.0407	28.9773	5100	0.3410	0.8976	0.7677	0.7865	0.7498
0.0445	29.5455	5200	0.3395	0.8969	0.7661	0.7851	0.7480

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0