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

# Experiment settings
exp_name: "brittle_star_production"
seed: 42

# Tracking settings - IMPORTANT: Set your own wandb_entity!
track: true
wandb_project_name: "PPO-Modularity"
wandb_entity: null  # ⚠️  SET THIS TO YOUR WANDB USERNAME OR TEAM

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

# Environment settings
num_envs: 32  # Increased for production

# Training hyperparameters - Production scale
total_timesteps: 50000000  # 50M timesteps for full training
learning_rate: 0.00025
num_steps: 256  # Longer rollouts
anneal_lr: true

# PPO specific - Fine-tuned
gamma: 0.99
gae_lambda: 0.95
num_minibatches: 8  # More minibatches for stability
update_epochs: 4
norm_adv: true
clip_coef: 0.2
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