kindredbluespirit/xvla-oct26-draft-model

Model Card for xvla

X-VLA is a soft-prompted, flow-matching Vision-Language-Action framework that treats each robot or hardware setup as a "task" encoded with a small set of learnable Soft Prompt embeddings, letting a single model reconcile diverse robot morphologies, sensors, and action spaces.

This policy has been trained and pushed to the Hub using LeRobot.

Learn how to train and run it in the LeRobot xvla guide, or browse the full documentation.

---

Model Details

• License: apache-2.0
• Fine-tuned from: lerobot/xvla-base
• Robot type: so100
• Cameras: laptop, phone

Inputs & Outputs

The policy consumes these observation features and produces these action features.

Inputs

Feature	Type	Shape
observation.images.image	VISUAL	(3, 256, 256)
observation.images.image2	VISUAL	(3, 256, 256)
observation.state	STATE	(8,)
observation.images.image3	VISUAL	(3, 224, 224)

Outputs

Feature	Type	Shape
action	ACTION	(6,)

Training Dataset

• Repository: kindredbluespirit/xvla-oct26-draft-dataset
• Episodes: 7783
• Frames: 4210444
• Frame rate: 30 FPS
• Task(s): "Grasp a battery and put it in the bin.", "Grasp a lego block and put it in the bin.", "pour a coffee into the cup", "Grab a bottle of soybeans and pour it into this cup.", "Grasp the block and drop it in the designated area.", "Grasp the block and place it in the designated area.", "Grasp a block and put it in the designated area.", "Grasp a block and put it in the designated area. ", "Pick up the oranage cylinder and place it in the circular cutout", "Pick up and place at grid spot 1.", "Pick up and place in grid spot 1.", "Pick up and place at grid spot 2.", "Pick up and place at grid spot 3.", "Pick up and place at grid spot 4.", "Pick up and place in grid spot 5.", "Pick up and place in grid spot 6.", "Pick up and place in grid spot 7.", "Pick up and place in grid spot 8.", "Pick up and place in grid spot 9.", "Pick up and place at grid spot 5", "Pick up and place at grid spot 6", "Pick up and place at grid spot 7", "Pick up and place at grid spot 8", "Grasp a dice and put it in the bin.", "Grasp a tape and put it in the bin.", "Move the red chess pieces as indicated by the highlighted squares", "Move the blue chess pieces as indicated by the highlighted squares", "Grasp a block and put it on board.", "Grasp the green cube and put it in the bin.", "Grasp the grey cube and put it in the bin.", "Pick up the charger and put it on the white tissue.", "Pick up the red cube and put it in the box.", "Pick up the red cube and stack it on the yellow cube.", "Grasp a pen and move it to a new platform.", "Grasp a medic box and put it in the bin.", "Grasp a yellow box and put it in the bin.", "Grasp a box and put it in the bin.", "Grasp candy to circle.", "Grasp candy to fix circle.", "Stack the red cube ontop of each other.", "Put candy into bowl.", "Grasp the object and put it on the cross", "Push a circle block to the blue square in the center.", "Sort rubbish by color.", "Pick up the yellow duck and put it in the box", "Pick up the black duck and put it in the box", "Grasp a yellow duck and put it in the box.", "Point First Point.", "Point First Point And Not Reset.", "Point It.", "Grasp a ball and put it on a goblet.", "Grasp the deck of cards and move it to the left.", "Grasp a chess and put it in the blue box.", "Grasp a chess and put it in blue box.", "Move the piece to destination", "Grasp a box and put it in the side.", "Grab the panda doll and put it in the bowl.", "Grasp a toxy and put it in the bin.", "Grasp a tape and put it in the drawing circle.", "shake advait's hand", "Pick up the cube and place it in the box.", "Remove the cube from the box.", "Pick up the pen and put it in the pen holder.", "Pick up the yellow tape and place it in the box.", "Remove the yellow tape from the box.", "Pick up the red cube and place it in the box.", "Grab up the yellow tape and put it in the box.", "Grasp a red block and put it in the bin.", "Grasp a black block and put it in the bin.", "Grasp a block and put it in the mug.", "Move box red to zone 1", "Move box red to zone 2", "Move box red to home", "Move box blue to zone 1", "Move box blue to zone 2", "Move box blue to home", "Move box yellow to zone 1", "Move box yellow to zone 2", "Move box yellow to home", "Grasp a cylinder and put it in the box", "Grasp the lego brick and put it in the box.", "Grasp the tape and put it in the bin.", "Grasp the yellow lego block and put it in the box.", "Grasp a yellow lego block and put it in the bin.", "Grasp the purple lego block and put it in the box.", "Grasp a lego block and put it in the box.", "サーボモーターの箱をシルバーの箱に入れる", "Grasp a tape reel and place it on the book", "Grasp a cube and put it in the box.", "Grasp a toy car and place it into the box.", "Grasp a small box and put it into the container.", "Grasp a taurus ring and put it in a glass cup.", "Grasp a Black stone and put it in the board on position (0,0).", "Grasp a block and put it in the bin.", "Grasp a wood brick and put it in the bin.", "Grasp a strawberry bear and put it in the bin.", "Grasp the red marker eraser and place it on the pink notebook.", "place blue block on red block, then place yellow block on blue block", "Pick up the blue block and place it on top of the red block.", "Put the red cube in the bin", "Pick up the red dice and place in bowl.", "grab the candle wick and place it in the tray.", "grab the candle wick by the aluminium plate and place it in the tray.", "Grasp the nerf bullet and put it in the bac.", "Grasp a green lego block and put it in the bin.", "Pick up the box.", "Grasp a skyblue doll and put it into the bowl.", "Grasp a mini car and put it in the bin.", "Grasp a bottle cap.", "Flip A0.", "flip A0", "Flip A1.", "flip A1", "Flip A2.", "flip A2", "Flip A3.", "flip A3", "Draw a circle.", "Pick up a new token and place it in the grid on square A1.", "Pick up the round, orange token and place it in the grid on square A1.", "Pick up the round, orange token and place it in the grid on square A2.", "Pick up the round, orange token and place it in the grid on square A3.", "Pick up a new token and plac it in the grid on square B1.", "Pick up the round, orange token and place it in the grid on square B1.", "Pick up the round, orange token and place it in the grid on square B2.", "Pick up the round, orange token and place it in the grid on square B3.", "Pick up a new token and plac it in the grid on square C1.", "Pick up the round, orange token and place it in the grid on square C1.", "Pick up a new token and plac it in the grid on square C2.", "Pick up the round, orange token and place it in the grid on square C2.", "Pick up a new token and plac it in the grid on square C3.", "Pick up the round, orange token and place it in the grid on square C3.", "Grasp a toy and put it in the bin.", "Grasp a ball and put it in the cup.", "Grasp a dice block and put it in the correct location.", "Grasp a paper box and put it in the bin.", "Pick a light wooden block and put it on board.", "Grasp a cate and put it in the box.", "Grasp a snack and put it on the plate.", "Go to the Red Jam.", "Place bagel on jig", "Grasp the bagel and put it in the center bin.", "Grasp a ball and put it in the box.", "Grasp a green lego block and put it in the red bin.", "Grasp a plug and put it in the bin.", "Grasp a hacky sack and put it in the bin.", "Grasp a lego and put it in the basket.", "Grasp a whale and put it in the plate.", "easy transfer", "Pass the marker.", "Hold this", "Hold", "Grab and Shake hand when you see it", "Grasp the tape and put it in the box.", "Grasp a dressing and put it in the cup.", "Grasp a sealant on left and put it to the right.", "Pick an Item out of the plage.", "Grasp a component and put it in the bin.", "Grasp a bacardi bottle and pour some into the cup."

Training Configuration

Setting	Value
Training steps	40000
Batch size	12
Optimizer	xvla-adamw
Learning rate	0.0001
Seed	1000
LeRobot version	0.6.1

---

How to Get Started with the Model

New to LeRobot? These guides cover the full workflow:

• Install LeRobot — set up the lerobot package.
• Hardware setup — assemble, wire, and calibrate your robot and cameras.
• Record data & train a policy — the end-to-end imitation-learning walkthrough.
• CLI cheat-sheet — quick reference for the lerobot-* commands.

The short version to run and train this policy:

Run the policy on your robot

lerobot-rollout \
  --strategy.type=base \
  --robot.type=so100 \
  --robot.port=<your_robot_port> \
  --robot.cameras="{ <camera_1>: {type: opencv, index_or_path: <index_or_path>, width: 640, height: 480, fps: 30}, <camera_2>: {type: opencv, index_or_path: <index_or_path>, width: 640, height: 480, fps: 30}}" \
  --policy.path=kindredbluespirit/xvla-oct26-draft-model \
  --task="Grasp a battery and put it in the bin." \
  --duration=60

Replace the remaining <...> placeholders with your own values: --robot.port and the camera names/indices are specific to your machine, and the camera names must match the observation keys this policy was trained on.

When --strategy.type=base is used the script doesn't record the episodes. Skipping duration will make the policy run indefinitely. For more information look at rollout documentation.

Train your own policy

This policy type is usually fine-tuned from the pretrained base model lerobot/xvla-base:

lerobot-train \
  --dataset.repo_id=${HF_USER}/<dataset> \
  --policy.path=lerobot/xvla-base \
  --output_dir=outputs/train/<policy_repo_id> \
  --job_name=lerobot_training \
  --policy.device=cuda \
  --policy.repo_id=${HF_USER}/<policy_repo_id> \
  --wandb.enable=true

Writes checkpoints to outputs/train/<policy_repo_id>/checkpoints/.

---

Evaluation

No evaluation results have been provided for this policy yet.

---

Citation

If you use this policy, please cite the method linked in the description above, along with LeRobot:

@misc{cadene2024lerobot,
    author = {Cadene, Remi and Alibert, Simon and Soare, Alexander and Gallouedec, Quentin and Zouitine, Adil and Palma, Steven and Kooijmans, Pepijn and Aractingi, Michel and Shukor, Mustafa and Aubakirova, Dana and Russi, Martino and Capuano, Francesco and Pascal, Caroline and Choghari, Jade and Moss, Jess and Wolf, Thomas},
    title = {LeRobot: State-of-the-art Machine Learning for Real-World Robotics in Pytorch},
    howpublished = "\url{https://github.com/huggingface/lerobot}",
    year = {2024}
}