Instructions to use Wjjjh/Recursive-VAM-models with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use Wjjjh/Recursive-VAM-models with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Wjjjh/Recursive-VAM-models", torch_dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
File size: 890 Bytes
d232921 6a16f4b d232921 6a16f4b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 | ---
license: apache-2.0
tags:
- robotics
- world-model
- recursive-search
- libero
---
# Recursive-VAM Task-3 models
Frozen model bundle for continuing the sparse-milestone and dynamic recursive
depth experiments in
[JunhaoWu7/Recursive-VAM](https://github.com/JunhaoWu7/Recursive-VAM).
Contents:
- `libero_goal_step5000_model/`: resolved transformer, text encoder, tokenizer,
and Wan VAE used by the frozen same-root D1/D2/D3 experiment;
- `action_semantic_goal_object_v3_seed42/`: frozen sparse cumulative milestone
head and its training summary.
The authoritative protocol, hashes, data paths, and interpretation rules are
in `docs/TASK3_HANDOFF_20260727.md` and
`experiments/task3_artifact_manifest.json` in the GitHub repository.
This repository contains model artifacts only. Matching preencoded data is in
the private dataset repository `Wjjjh/Recursive-VAM-data`.
|