chessdata-model

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: 0.5827
• Accuracy: 0.8378

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	7	1.1069	0.7207
1.0143	2.0	14	1.0853	0.7117
0.9148	3.0	21	0.9472	0.7297
0.9148	4.0	28	0.8859	0.7568
0.7721	5.0	35	0.8500	0.7658
0.71	6.0	42	0.7973	0.8108
0.71	7.0	49	0.8040	0.7748
0.641	8.0	56	0.8344	0.7207
0.6122	9.0	63	0.7528	0.7748
0.5698	10.0	70	0.8087	0.7748
0.5698	11.0	77	0.7347	0.7838
0.5329	12.0	84	0.6237	0.8288
0.5264	13.0	91	0.6135	0.8378
0.5264	14.0	98	0.7670	0.7568
0.4846	15.0	105	0.6465	0.8288
0.4597	16.0	112	0.6354	0.8288
0.4597	17.0	119	0.7096	0.7838
0.409	18.0	126	0.6364	0.8468
0.4321	19.0	133	0.6343	0.8108
0.4309	20.0	140	0.5827	0.8378

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2