vit-base-hate-meme

This model is a fine-tuned version of google/vit-base-patch16-224 on the emily49/hateful_memes_train_dev dataset. It achieves the following results on the evaluation set:

• Loss: 0.6966

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: 0.0002
• train_batch_size: 16
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 8
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.6407	1.0	532	0.6966
0.5847	2.0	1064	0.7980
0.6144	3.0	1596	0.7764
0.5307	4.0	2128	0.9913
0.5545	5.0	2660	0.9512
0.3243	6.0	3192	1.5071
0.1371	7.0	3724	2.0203
0.0571	8.0	4256	2.9499

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2