tdpeuter/2026SEL3-project-Brittle_Star_Controller_Modularity
tdpeuter/2026SEL3-project-Brittle_Star_Controller_Modularity
Archived
1
Fork 0
Code Issues Releases Activity

feat(logging): added amount of envs that truncated and terminated

Browse source
This commit is contained in:
vsc46589 vscuser 2026-04-09 23:40:58 +02:00
parent b839024e6e
commit f7428f9d90
1 changed files with 13 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

14
src/brittle_star_project/trainers/PPOTrainer.py
View file

@ -203,6 +203,8 @@ class LossInfo:
avg_episodic_return: Any
# Example of better typing
explained_variance: float
num_terminated: int
num_truncated: int
class PPOTrainer:
@ -359,6 +361,8 @@ class PPOTrainer:
.hyperparams["learning_rate"]
.item(),
"charts/explained_variance": loss_info.explained_variance,
"charts/num_terminated": loss_info.num_terminated,
"charts/num_truncated": loss_info.num_truncated,
"losses/value_loss": loss_info.v_loss[-1, -1].item(),
"losses/policy_loss": loss_info.pg_loss[-1, -1].item(),
"losses/entropy": loss_info.entropy_loss[-1, -1].item(),
@ -408,6 +412,12 @@ class PPOTrainer:
storage.values, storage.returns
)
terminated = next_env_state.terminated # (num_envs,)
truncated = next_env_state.truncated # (num_envs,)
num_terminated = int(jnp.sum(terminated).item())
num_truncated = int(jnp.sum(truncated).item())
return (
next_env_state,
next_obs,
@ -419,7 +429,9 @@ class PPOTrainer:
entropy_loss=entropy_loss,
approx_kl=approx_kl,
avg_episodic_return=avg_episodic_return,
explained_variance=explained_var
explained_variance=explained_var,
num_terminated=num_terminated,
num_truncated=num_truncated
),
)