Skip to content
Robotics

VivekSai07/xvla_hfteam_box_middle

fine-tune for robotic arms

2 min read

The X-VLA model (0.88B) has been fine-tuned on teleoperation demonstrations for synchronized dual-arm operation using SO-101 robotic arms. The architecture processes data from three cameras, 12 joint positions, and a text prompt, outputting control actions. This fine-tune (adapting model weights for a specific task) was created during a hackathon for cooperatively lifting a box and moving it to the center. The solution is useful for engineers and researchers working with the LeRobot library.

Key features

  • Base model: lerobot/xvla-base (0.88 billion parameters).
  • Input data: three cameras, 12 joint positions, and task text.
  • Control: generation of 1-second action chunks for two robotic arms.
  • Training: 7,000 steps with a batch size of 16 and bf16 autocast (automatic mixed precision data type).

What you can do

  • Evaluate the model's behavior on dual SO-101 robotic arms.
  • Test cooperative box transfer via a text instruction.
  • Apply the weights for multi-arm system experiments.

Example

lerobot-rollout --strategy.type=base \
    --policy.path=VivekSai07/xvla_hfteam_box_middle \
    --robot.type=bi_so_follower \
    --task="Lift the box with both arms and place it in the middle" --duration=30
Read moreDetails
Training: 41 teleoperation episodes (20,556 frames at 30 fps).
Metrics: training loss dropped from 0.79 to 0.003 at step 7,000.
Quality: average error on training frames is 0.9 degrees.
Limitations: success rate on real hardware has not been measured yet.
Discuss Telegram 0

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 3.3 GB · ready files4.1 GB
  • BF16weights 1.6 GB2.3 GB
  • FP8 / INT8recommendedweights 0.8 GB1.4 GB
  • 4-bitweights 0.5 GB1.0 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: FP8 / INT8 — from RTX 2060 · F32 — from RTX 2060

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.

    # VivekSai07/xvla_hfteam_box_middle
    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 VivekSai07/xvla_hfteam_box_middle --local-dir ~/models/xvla_hfteam_box_middle
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to xvla_hfteam_box_middle
  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 xvla-base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 1All on Hugging Face
Fine-tunes 12All on Hugging Face

About the model

VivekSai07/xvla_hfteam_box_middle
Author
VivekSai07
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments