# Production Training Configuration
#
# Full-scale training configuration for production runs
# with wandb logging enabled.

# Experiment settings
exp_name: "brittle_star_production_training"
seed: 42

# Tracking
track: true
capture_video: false
wandb_project_name: "PPO-Modularity"
wandb_entity: "SEL3-2026-Groep-4"

# Model saving
save_model: true
checkpoint_frequency: 100  # Save checkpoint every 100 iterations

# Environment settings
num_envs: 512

# Training hyperparameters
total_timesteps: 50000000
num_steps: 256
num_minibatches: 4
update_epochs: 4

learning_rate: 2.5e-4
anneal_lr: true
gamma: 0.99
gae_lambda: 0.95
clip_coef: 0.1
clip_vloss: true
ent_coef: 0.01
vf_coef: 0.5
max_grad_norm: 0.5
target_kl: null

# Hardware
cuda: true
torch_deterministic: true