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]Recursive-VAM Task-3 models
Frozen model bundle for continuing the sparse-milestone and dynamic recursive depth experiments in 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.
- Downloads last month
- -