Downloads · 30 days
2
18% of all-time downloads
ljw07/openwam-libero-isaacsim-step128000-resume
openwam-libero-isaacsim-step128000-resume is a robotics model from ljw07. Use it for the robotics task on the model card, and read the license before you ship it in a product. It is set up for openwam. The card lists the license as apache-2.0.
This repository preserves the exact DeepSpeed/Accelerate continuation state for the local OpenWAM-Alpha LIBERO IsaacSim baseline paused on 2026-09-16.
Downloads · 30 days
2
18% of all-time downloads
All-time downloads
11
Public
Repo size
135 GB
Likes
0
Public
Click a slice to open those files.
.pt110 GB · 81%
From the Hugging Face model README
This repository preserves the exact DeepSpeed/Accelerate continuation state for the local OpenWAM-Alpha LIBERO IsaacSim baseline paused on 2026-09-16.
128000 / 170420128656accel_state_step_128000/ contains the model, optimizer, scheduler, RNG, and
trainer cursor state. Exact continuation requires the same world size of 8.
See TRAINING_HANDOFF_STEP128000.md for paths, caveats, and the prepared
launcher.
The 24.8 GB checkpoint_step_128000.safetensors is included as a portable
warm-start checkpoint for future machines whose GPU count differs from the
original eight-GPU topology. The full DeepSpeed state is also included for
exact continuation.
hf download ljw07/openwam-libero-isaacsim-step128000-resume \
--local-dir /path/to/2026-09-14_23-38-07
Place the downloaded directory at the run-state path expected by
scripts/resume_openwam_alpha_libero_isaacsim_step128000.sh. The OpenWAM code,
adapted dataset, and Python environment are separate from this checkpoint.
This run uses a local ModelScope IsaacSim LIBERO adaptation. Its actions are next-pose-derived EEF10 proxies rather than verified simulator-native commands. Training loss from this run is not an official rollout success rate.