CanIEmbody
Matrix/Compute/CPU only

What can I run on CPU only (no accelerator)?

system RAMView matrix for this target →

memory_gb is 0 because there is no dedicated accelerator memory: models run from system RAM, which varies per machine. CPU model, core count and RAM change results by large factors, so runs on this target must record the CPU in platform.cpu (or platform.chip) and the RAM size.

Policy by policy

PolicyParamsInference memory (sourced, with precision)On CPU only (no accelerator)Best measured
ACT Imitation80M211 MB measured fp32no data
Diffusion Policy Imitation—1.12 GB measured fp32no data
EO-1 VLA3B≥6.5 GBno data
Evo-1 VLA770M—no data
Fast-WAM World model6B—no data
GR00T N1.7 VLA3B≥16 GBno data
HIL-SERL RL——no data
LingBot-VA World model5.3B≥18 GBno data
MolmoAct2 VLA—12.1 GB measured bf16no data
Multitask DiT Imitation450M—no data
OpenVLA VLA7.5B15 GB measured bf16no data
OpenVLA-OFT VLA—≥16 GBno data
SmolVLA VLA450M1.75 GB measured fp32no data
TD-MPC RL——no data
VLA-JEPA VLA——no data
VQ-BeT Imitation——no data
WALL-OSS VLA——no data
X-VLA VLA900M—no data
π0 VLA3.3B13.32 GB measured fp32no data
π0-FAST VLA—≥8 GBno data
π0.5 VLA—≥8 GBno data

“Exceeds memory” compares a policy's upstream-stated minimum inference memory (never a measured peak, which is precision-specific) with this target's VRAM. It is arithmetic, not observation, and is labelled inferred. Quantised or split runtimes can change the answer; a measured run always wins.

All measured runs on CPU only (no accelerator)

No measured runs on CPU only (no accelerator) yet. A run needs runtime, precision, camera count. Contribute one.

Sources

  1. T1 · upstream documentedofficial-docssource 2026-08-03verified 2026-10-24 (1d ago)lerobot 0.6.1

    LeRobot docs list CPU as a backend where all computation runs on the CPU with no accelerator.

    “- **CPU**: CPU executes all computations, no dedicated accelerator is used”
    https://github.com/huggingface/lerobot/blob/v0.6.1/docs/source/torch_accelerators.mdx