akoulapure's picture
Initial Deployment: Best ViT Model
6cc8ae1 verified
Raw
History Blame Contribute Delete
400 Bytes
# MLP Baseline Configuration
# Inherits from base.yaml
model:
name: "mlp"
architecture:
hidden_layers: [1024, 512, 256]
dropout: 0.3
batch_norm: true
activation: "relu"
training:
batch_size: 64
num_epochs: 50
learning_rate: 0.001
weight_decay: 0.0005
optimizer: "adam"
scheduler: "step"
scheduler_step_size: 15
scheduler_gamma: 0.1
early_stopping_patience: 10