Skip to content
Robotics

PerceptronAI/Isaac-0.5-YAM

a policy for dual-arm robots

2 min read

This new model is a fine-tuned version of Isaac 0.5 trained on dual-arm YAM demonstrations from xDoF ABC and MolmoAct2. The policy takes inputs from three RGB cameras, a text instruction, and the current joint states of the manipulators, then predicts target joint and gripper values for both arms. The checkpoint was saved after 200,000 steps of fine-tuning across 32 GPUs. It is built for robotics developers and researchers working with the lerobot library.

Key features

  • Fine-tuned on dual-arm YAM demonstrations (xDoF ABC and MolmoAct2).
  • Processes three RGB camera feeds (a wide-angle view and two wrist-mounted cameras).
  • Predicts absolute target joint and gripper values for both arms.
  • Outputs an action chunk of size 60 by 14 for each observation.

What you can do

  • Run policy inference on custom or recorded observations using Python.
  • Fine-tune the model on a custom LeRobot dataset using two GPUs with CPU offload.
  • Integrate the model into projects involving dual-arm robotic manipulator control.

Example

import json
from pathlib import Path
import torch
from lerobot.policies.factory import make_pre_post_processors
from lerobot.policies.perceptron_isaac import PerceptronIsaacPolicy
from torchvision.io import decode_image

snapshot = Path("./Isaac-0.5-YAM")
assets = snapshot / "assets"
sample = json.loads((assets / "sample_observation.json").read_text())
Read moreDetails
Takes three RGB cameras, arm states, and a task prompt
Outputs an action chunk of 60 by 14
Trained on 32 GPUs with a global batch size of 256 sequences
Examples
Discuss Telegram 1 2

Will it run?

Only with RAM offload

Even 4-bit is larger than your VRAM. Some layers can stay in system RAM (llama.cpp, ComfyUI), but it will be several times slower.

  • BF16weights 133 GB · ready files147 GB
  • FP8 / INT8recommendedweights 67 GB74 GB
  • 4-bitweights 37 GB42 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: Won't fit any home GPU — renting is cheaper.

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # PerceptronAI/Isaac-0.5-YAM
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download PerceptronAI/Isaac-0.5-YAM --local-dir ~/models/Isaac-0.5-YAM
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Isaac-0.5-YAM
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of Isaac-0.5-YAM (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

PerceptronAI/Isaac-0.5-YAM
Author
PerceptronAI
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

21

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments