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

feat: log distance to target for each env

Browse source
This commit is contained in:
JibrilExe 2026-04-12 13:09:53 +02:00
parent c5a8dd2b3a
commit 631d01ebf6
2 changed files with 7 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

3
configs/hpc/debug.yaml
View file

@ -7,8 +7,7 @@ wandb_entity: "SEL3-2026-Groep-4"
run_dir: "/data/gent/465/vsc46589" run_dir: "/data/gent/465/vsc46589"
num_envs: 32 num_envs: 32
num_steps: 32 num_steps: 32
total_timesteps: 102400 total_timesteps: 409600
num_arms: 2 num_arms: 2
num_segments_per_arm: 1
cuda: true cuda: true

12
src/brittle_star_project/trainers/PPOTrainer.py
View file

@ -407,10 +407,10 @@ class PPOTrainer:
xy_distance xy_distance
): ):
data = jax.device_get({ data = jax.device_get({
'rewards': storage.rewards[0], # (num_steps,) 'rewards': storage.rewards[0],
'values': storage.values[0], 'values': storage.values[0],
'returns': storage.returns[0], # NEW 'returns': storage.returns[0],
'advantages': storage.advantages[0],# NEW 'advantages': storage.advantages[0],
'actions': storage.actions[0], 'actions': storage.actions[0],
'raw_actions': storage.raw_actions[0], 'raw_actions': storage.raw_actions[0],
'means': storage.means[0], 'means': storage.means[0],
@ -433,11 +433,11 @@ class PPOTrainer:
"rollout/env0/action_mean": float(np.mean(data['actions'])), "rollout/env0/action_mean": float(np.mean(data['actions'])),
"rollout/env0/raw_action_mean": float(np.mean(data['raw_actions'])), "rollout/env0/raw_action_mean": float(np.mean(data['raw_actions'])),
"charts/env0_xy_distance_to_target": float(xy_distance[0]),
"charts/env1_xy_distance_to_target": float(xy_distance[1]),
} }
for i in range(len(xy_distance)):
storage_metrics[f"env_data/env{i}_xy_dist_target"] = float(xy_distance[i])
metrics = { metrics = {
"charts/avg_episodic_return": training_measurements.avg_episodic_return, "charts/avg_episodic_return": training_measurements.avg_episodic_return,
"charts/avg_episodic_length": np.mean( "charts/avg_episodic_length": np.mean(