# PPO Production Training Profile

[training]
epochs = 100
batch_size = 64
learning_rate = 0.0001
actor_learning_rate = 0.000001
critic_learning_rate = 0.001
gamma = 0.99

[ppo]
clip_epsilon = 0.2
value_loss_coef = 1.0
entropy_coef = 0.05
gae_lambda = 0.95
rollout_steps = 2048
minibatch_size = 64
max_grad_norm = 0.5

[early_stopping]
enabled = true
patience = 30
min_epochs_before_stopping = 50
min_value_loss_improvement_pct = 2.0

[experience]
initial_capital = 1000000.0
gpu_n_episodes = 128
gpu_timesteps_per_episode = 500
avg_spread = 0.0001
accumulation_steps = 1