aoi_clip_high_resolution_cross_attention_fusin

This model is a fine-tuned version of OFA-Sys/chinese-clip-vit-base-patch16 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 4.0266
• Accuracy: 0.0266

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: 1e-05
• train_batch_size: 40
• eval_batch_size: 40
• seed: 42
• gradient_accumulation_steps: 5
• total_train_batch_size: 200
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 60.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.4856	5.9923	1866	3.6872	0.0237
2.4767	11.9846	3732	3.6874	0.0244
2.4784	17.9769	5598	3.6876	0.0247
2.4728	23.9692	7464	3.6878	0.0251
2.473	29.9615	9330	3.6875	0.0255
2.4621	35.9538	11196	3.7371	0.0266
2.3868	41.9461	13062	3.8208	0.0269
2.3293	47.9383	14928	3.8805	0.0268
2.2917	53.9306	16794	3.9713	0.0267
2.2787	59.9229	18660	4.0266	0.0266

Framework versions

• Transformers 4.42.3
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1