document-spoof

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: 0.1105
• Accuracy: 0.9767

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	0.9524	5	0.5211	0.8837
No log	1.9048	10	0.2271	0.8837
0.545	2.8571	15	0.0975	0.9884
0.545	4.0	21	0.1020	0.9767
0.545	4.9524	26	0.3087	0.9535
0.472	5.9048	31	0.3385	0.8023
0.472	6.8571	36	0.2358	0.8605
0.472	8.0	42	0.3675	0.8605
0.3762	8.9524	47	0.1460	0.9535
0.3762	9.9048	52	0.6158	0.8140
0.3762	10.8571	57	0.3228	0.9186
0.1586	12.0	63	0.0248	0.9884
0.1586	12.9524	68	0.0639	0.9651
0.1586	13.9048	73	0.5674	0.8488
0.1159	14.8571	78	0.0291	0.9884
0.1159	16.0	84	0.0539	0.9884
0.1159	16.9524	89	0.0772	0.9767
0.0366	17.9048	94	0.0031	1.0
0.0366	18.8571	99	0.1506	0.9535
0.0179	20.0	105	0.0007	1.0
0.0179	20.9524	110	0.1427	0.9535
0.0179	21.9048	115	0.2299	0.9419
0.0036	22.8571	120	0.1373	0.9767
0.0036	23.8095	125	0.1105	0.9767

Framework versions

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