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

fix(ppo training): forgot extra change from dev

Browse source
This commit is contained in:
Robin Meersman 2026-05-07 08:42:35 +02:00
parent 090e774ad7
commit 23fcee2110
1 changed files with 8 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

13
src/brittle_star_project/trainers/PPOTrainer.py
View file

@ -129,18 +129,21 @@ def _step_once(
logger11.debug(f"[_step_once] value: {value.shape}")
logger11.debug(f"[_step_once] mean: {mean.shape}")
logger11.debug(f"[_step_once] std: {std.shape}")
key, reset_key = jax.random.split(key)
reset_rngs = jax.random.split(reset_key, num_envs)
# ---- ENV STEP ----
key, reset_key = jax.random.split(key)
reset_rngs = jax.random.split(reset_key, num_envs)
episode_stats, env_state, (next_obs, reward, next_done) = env_step_fn(
episode_stats, env_state, flat_clipped_action, reset_rngs,
episode_stats, env_state, (next_obs, reward, next_done, terminated, truncated) = env_step_fn(
episode_stats,
env_state,
flat_clipped_action,
reset_rngs,
)
terminated_any = terminated_any | terminated
truncated_any = truncated_any | truncated