Repository
Open source
LeRobot · huggingface
LeRobot tools for robot control, datasets and end-to-end learning.
Robot learning frameworks
Models / Research
SimpleMemVLA · OpenBMB
Native-video memory for vision-language-action models, using timestamped visual history and exact streaming inference for long-horizon robot manipulation.
Overview
Native-video memory for vision-language-action models, using timestamped visual history and exact streaming inference for long-horizon robot manipulation.
Review observation/action interfaces, checkpoints, robot embodiments and adaptation requirements. Repository code licensing does not automatically license weights or training data.
Project information
- Maintainer
- OpenBMB
- Primary language
- Python
- Repository code license
- MIT
- Model-weight license
- Not recorded
- Dataset license
- Not recorded
- Hardware design license
- Not recorded
- Maintenance
- Not archived in the last recorded metadata
- Last observed repository push
- 2026-09-24
Code licenses do not automatically cover model weights, datasets or hardware designs. Unrecorded license details require confirmation with the publisher.
Documentation sections
- 1) core env (repeat per benchmark, e.g. simplememvla-rmbench, simplememvla-libero, ...)
- 2) fast-path kernels: flash-attn (required — the default attention backend for
- train AND eval) + flash-linear-attention + causal-conv1d (required for training,
- optional for eval). Needs nvcc for causal-conv1d.
- 3) the benchmark's simulator stack
- SAPIEN headless rendering additionally needs the Vulkan loader:
