ktp-not-ktp-clip

This model is a fine-tuned version of openai/clip-vit-base-patch32 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.1267
• Accuracy: 0.9890

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: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	7	0.5809	0.6374
No log	2.0	14	1.3401	0.6703
0.5558	3.0	21	0.6458	0.7692
0.5558	4.0	28	0.3785	0.8681
0.1701	5.0	35	0.3004	0.9451
0.1701	6.0	42	0.2204	0.9560
0.142	7.0	49	0.1483	0.9341
0.142	8.0	56	0.1386	0.9670
0.1002	9.0	63	0.7714	0.8681
0.1002	10.0	70	0.2285	0.9341
0.0956	11.0	77	0.1162	0.9780
0.0956	12.0	84	0.1104	0.9780
0.0004	13.0	91	0.1722	0.9780
0.0004	14.0	98	0.2109	0.9780
0.0209	15.0	105	0.3321	0.9560
0.0209	16.0	112	0.0785	0.9780
0.0209	17.0	119	0.1525	0.9670
0.0014	18.0	126	0.1436	0.9670
0.0014	19.0	133	0.2278	0.9670
0.0002	20.0	140	0.3035	0.9560
0.0002	21.0	147	0.1239	0.9780
0.001	22.0	154	0.1211	0.9890
0.001	23.0	161	0.1253	0.9890
0.0	24.0	168	0.1265	0.9890
0.0	25.0	175	0.1267	0.9890

Framework versions

• Transformers 4.41.2
• Pytorch 2.1.2
• Datasets 2.19.2
• Tokenizers 0.19.1