agent.resume param isn't work

[heptarri]

When I interrupted a tranning session and use --agent.resume True to restart it, it starts a new session. Is that a bug or only because I don't know how to use it?

my script:

python scripts/train.py Unitree-G1-Tracking --motion_file src/assets/motions/g1/G1.npz --agent.resume True

[psh1234567849]

add this code in train.py:
##################
if resume_path is not None:
print(f"[INFO] Loading checkpoint from: {resume_path}")
runner.load(str(resume_path), map_location=device)
############
before runner.learn(
num_learning_iterations=cfg.agent.max_iterations, init_at_random_ep_len=True
)
and perhaps you need more detailed param like this:
python scripts/train.py Unitree-G1-Flat --env.scene.num-envs=4096 --agent.resume=True --agent.load-run="2026-03-10_22-06-00" --agent.load-checkpoint="model_500.pt"

[swumh]

When I interrupted a tranning session and use --agent.resume True to restart it, it starts a new session. Is that a bug or only because I don't know how to use it?

my script:

python scripts/train.py Unitree-G1-Tracking --motion_file src/assets/motions/g1/G1.npz --agent.resume True

Hi, I believe this issue has been resolved in my PR #18.
You can check the details here: #18
Please feel free to check it out. Any feedback is welcome.

usage:

python scripts/train.py Unitree-G1-Flat   --agent.resume=True   --agent.load_run=2026-03-10_18-13-38   --agent.load_checkpoint=model_9500.pt  --env.scene.num-envs=4096