import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("dsadsdasd/ours", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]Ours: per-image Turtle + InternVLA full-32 /3 + Full48
This is the complete weight bundle for the locked current Ours protocol. It includes one standard Hugging Face navigation model containing the final V3 InternVLA System2 weights and the Turtle System1 trajectory DiT, action decoder, and e0.25 action encoder. There is no duplicated InternVLA base/overlay pair or separate Turtle weight file. The bundle also includes the FlashWorld runtime with the Full48 specialist DiT. The obsolete DiT embedded in the Joint e0.25 training checkpoint is not included.
The model card intentionally reports no formal benchmark numbers yet. Existing 26-scene outputs are protocol-audit artifacts, not formal Table 1 results. See ours_config.json for the exact component identities and construction protocol.
- Downloads last month
- -
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support