Keyurjotaniya007/microsoft-swin-base-img-emotion

swin-base

This model is a fine-tuned version of microsoft/swin-base-patch4-window7-224 on a Keyurjotaniya007/img-emotion-classification dataset. It achieves the following results on the evaluation set:

• Eval loss: 0.8799635767936707

• Eval Accuracy: 0.6767619586753467

Training hyperparameters

The following hyperparameters were used during training:

• remove_unused_columns=False

• learning_rate=3e-6

• per_device_train_batch_size=32

• gradient_accumulation_steps=4

• per_device_eval_batch_size=32

• num_train_epochs=25

• warmup_steps=50

• weight_decay=0.02

• metric_for_best_model="accuracy"

• fp16=True

• lr_scheduler_type= "cosine"