vit-letter-identification-v3

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

• Loss: 1.3440
• Accuracy: 0.7615

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: 2e-05
• train_batch_size: 80
• eval_batch_size: 80
• seed: 1337
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 250.0

Training results

Training Loss	Epoch	Step	Accuracy	Validation Loss
No log	1.0	7	0.0154	3.9449
3.9333	2.0	14	0.0231	3.9367
3.8939	3.0	21	0.0308	3.9280
3.8939	4.0	28	0.0462	3.9167
3.8562	5.0	35	0.0692	3.9033
3.8008	6.0	42	0.0769	3.8874
3.8008	7.0	49	0.1077	3.8670
3.7555	8.0	56	0.1	3.8495
3.6917	9.0	63	0.1154	3.8305
3.6372	10.0	70	0.1385	3.8138
3.6372	11.0	77	0.1231	3.7966
3.5846	12.0	84	0.1538	3.7767
3.5047	13.0	91	0.2308	3.7516
3.5047	14.0	98	0.2385	3.7279
3.4547	15.0	105	0.2385	3.7031
3.3796	16.0	112	0.2692	3.6725
3.3796	17.0	119	0.2769	3.6462
3.3283	18.0	126	0.2923	3.6226
3.2728	19.0	133	0.2846	3.6022
3.2229	20.0	140	0.2769	3.5930
3.2229	21.0	147	0.3308	3.5748
3.1514	22.0	154	0.3385	3.5404
3.1179	23.0	161	0.3385	3.5146
3.1179	24.0	168	0.3462	3.4916
3.0559	25.0	175	0.3385	3.4733
3.0051	26.0	182	0.3615	3.4540
3.0051	27.0	189	0.3692	3.4499
2.9775	28.0	196	0.3769	3.4355
2.9277	29.0	203	0.3846	3.4166
2.9066	30.0	210	0.4	3.4007
2.9066	31.0	217	0.3692	3.3826
2.8464	32.0	224	0.4077	3.3698
2.8044	33.0	231	0.4077	3.3509
2.8044	34.0	238	0.3769	3.3243
2.7699	35.0	245	0.3923	3.3201
2.7251	36.0	252	0.4	3.3013
2.7251	37.0	259	0.4231	3.2936
2.6915	38.0	266	0.4538	3.2827
2.6527	39.0	273	0.4615	3.2627
2.6541	40.0	280	0.4615	3.2581
2.6541	41.0	287	0.4231	3.2342
2.5968	42.0	294	0.4385	3.2211
2.573	43.0	301	0.4077	3.2122
2.573	44.0	308	0.4615	3.2259
2.554	45.0	315	0.4308	3.2271
2.5222	46.0	322	0.4462	3.2208
2.5222	47.0	329	0.4462	3.2139
2.5085	48.0	336	0.4538	3.2040
2.4593	49.0	343	0.4923	3.2053
2.4585	50.0	350	0.4769	3.1822
2.4585	51.0	357	0.4692	3.1697
2.4228	52.0	364	0.4692	3.1589
2.3954	53.0	371	0.4769	3.1375
2.3954	54.0	378	0.4538	3.1092
2.3641	55.0	385	0.4769	3.0999
2.3651	56.0	392	0.4615	3.0860
2.3651	57.0	399	0.4615	3.0813
2.3182	58.0	406	0.4923	3.0692
2.3029	59.0	413	0.4846	3.0610
2.2988	60.0	420	0.4615	3.0627
2.2988	61.0	427	0.4692	3.0520
2.2865	62.0	434	0.4538	3.0395
2.2623	63.0	441	0.4615	3.0357
2.2623	64.0	448	0.4615	3.0333
2.2252	65.0	455	0.4769	3.0229
2.2339	66.0	462	0.4769	3.0203
2.2339	67.0	469	0.4923	3.0076
2.2017	68.0	476	0.4846	2.9876
2.1972	69.0	483	0.4923	2.9716
2.1964	70.0	490	0.5	2.9632
2.1964	71.0	497	0.4923	2.9597
2.1775	72.0	504	0.5	2.9581
2.1619	73.0	511	0.5077	2.9516
2.1619	74.0	518	0.5154	2.9356
2.1633	75.0	525	0.5077	2.9286
2.1207	76.0	532	0.5154	2.9266
2.1207	77.0	539	0.5231	2.9205
2.1353	78.0	546	0.5154	2.9131
2.1075	79.0	553	0.5231	2.9075
2.1025	80.0	560	0.5231	2.9073
2.1025	81.0	567	0.5154	2.9174
2.1031	82.0	574	0.5308	2.9131
2.0932	83.0	581	0.5308	2.9092
2.0932	84.0	588	0.5308	2.8978
2.0861	85.0	595	0.5308	2.8871
2.0478	86.0	602	0.5385	2.8829
2.0478	87.0	609	0.5462	2.8804
2.0815	88.0	616	0.5462	2.8725
2.0756	89.0	623	0.5462	2.8694
2.065	90.0	630	0.5462	2.8665
2.065	91.0	637	0.5462	2.8615
2.0572	92.0	644	0.5462	2.8599
2.0358	93.0	651	0.5462	2.8620
2.0358	94.0	658	0.5462	2.8629
2.0663	95.0	665	0.5538	2.8625
2.0353	96.0	672	0.5538	2.8628
2.0353	97.0	679	0.5538	2.8629
2.0506	98.0	686	0.5538	2.8622
2.0494	99.0	693	0.5538	2.8622
2.0566	100.0	700	0.5538	2.8622

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.4.0
• Tokenizers 0.15.0