Skip to content
Robotics

SmolVLA for SO-101 Pouring

bimanual robot control fine-tune

1 min readupdated

This is a fine-tuned version of the 0.45B-parameter lerobot/smolvla_base model built to control a bimanual SO-101 robot. It was trained on demonstrations of pouring liquids and granular materials. Taking feeds from three cameras alongside 12 joint states, the model outputs control actions for both robotic arms. It is aimed at robotics researchers and developers working with the lerobot framework.

Key features

  • Built on lerobot/smolvla_base (0.45B parameters).
  • Uses three high-resolution cameras (top-down view, left wrist, and right wrist).
  • Supports bimanual control with 12-dimensional state and action spaces.
  • Trained on 136 demonstration episodes (176,654 frames).

What you can do

  • Automate liquid and powder pouring tasks on an SO-101 robotics setup.
  • Trigger specific robot actions using text prompts, such as pouring water or powders.
  • Drop the checkpoint into your own lerobot experiments to test bimanual control policies.

Example

Use the following text prompts to trigger specific tasks:
- Pour water into the cup. (right arm)
- Pour green tea latte powder from the green-lidded container into the cup. (left arm)

Read moreDetails
Training: 40,000 steps with a batch size of 24.
Validation loss dropped from 0.101 to 0.0457.
These metrics reflect imitation loss rather than real-world task success rates on physical hardware.
Discuss Telegram 0 25

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.8 GB · ready files1.4 GB
  • FP8 / INT8recommendedweights 0.4 GB1.0 GB
  • 4-bitweights 0.2 GB0.8 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 · BF16 — 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.

    # tkpota/smolvla_so101_pour
    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 tkpota/smolvla_so101_pour --local-dir ~/models/smolvla_so101_pour
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to smolvla_so101_pour
  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

  1. SmolLM2-360Mfamily root
  2. SmolLM2-360M-Instruct
  3. SmolVLM-500M-Instruct
  4. SmolVLM2-500M-Video-Instruct
  5. smolvla_base
  6. smolvla_so101_pour

New posts about versions of SmolLM2-360M (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 8All on Hugging Face
Merges 1All on Hugging Face

About the model

tkpota/smolvla_so101_pour
Author
tkpota
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

250

Ask AI about this model

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

Comments

Robotics

Jevol-2B

Jevol-2B