H2walk-main-30k

IsaacLab / RSL-RL policy checkpoints exported from 2026-10-22_04-45-07.

• checkpoints: 16 (model_0 → model_29999)
• total size: 270 MB
• params/env.yaml / params/agent.yaml: the exact environment and PPO configuration used for this run (frozen at launch time).

Loading

import torch
blob = torch.load('model_29999.pt', map_location='cpu')
actor_state = {k[6:]: v for k, v in blob['model_state_dict'].items()
               if k.startswith('actor.') and k != 'actor.std'}
norm = blob['obs_norm_state_dict']   # running mean/std of the observations

Every *.pt also contains critic.*, the observation normaliser and the iteration counter, so training can be resumed with --resume.