Greenwich

lloydlei/Greenwich

Greenwich is a robotics model from lloydlei. Use it for the robotics task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.

Weights + atlas for the AlphaMotion engine (https://github.com/Lloyd-lei/Greenwich):

Downloads · 30 days

Access

Public

Updated Sep 10, 2026

Repo size

1.3 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.npy638 MB · 50%

At a glance

Task: Robotics
License: apache-2.0
Access: Public
Created: Aug 14, 2026
Updated: Sep 10, 2026
SHA: b8b5b673

Model card

From the Hugging Face model README

AlphaMotion release artifacts (v0.2, pre-release)

Weights + atlas for the AlphaMotion engine (https://github.com/Lloyd-lei/Greenwich):

artifact	what
`greenwich/`	v0.1 cross-embodiment spatial codec (18.1M params, repaired decoder: round-trip fidelity 0.61 vs 0.11 pre-repair); code space of the Atlas, library and Equator
`bumblebee/`	v0.2 Greenwich-v2 codec, size S (3.7M, d128, 3+3 layers) — 132×13 codes, contact channel, root-height slot
`megatron/`	v0.2 Greenwich-v2 codec, size M (18.0M, d256, 4+4 layers) — recommended default
`prime/`	v0.2 Greenwich-v2 codec, size L (57.7M, d384, 6+6 layers)
`equator_a/`	temporal codec: 60f code window ↔ 32 tokens (40M)
`equator_b/`	bridge prior P(interior \| start, goal, n) (12.9M)
`atlas/`	65,536-window rainbow-code index (the Atlas Map)
`library/`	4,096 fully-regenerable clips (tokens + boundary codes, 23 MB)
`embodiments/`	Qwen3 joint-name embedding cache (299 names)

Install the engine, then alphamotion download pulls the v0.1 stack plus megatron; alphamotion download --models bumblebee,megatron,prime pulls the whole v2 ladder. The v2 codecs need the embodiments_v2 descriptor set that ships inside the package (not here). Audit numbers: docs/BENCHMARK.md. Benchmark: see docs/BENCHMARK.md in the code repo; reproduce with alphamotion eval.

Task: Robotics
License: apache-2.0
Created: Aug 14, 2026
Updated: Sep 10, 2026