Model Card for z18820636149/SmolVLA_pick_and_place_v2

This repository contains a smolvla policy exported from SparkMind.

Summary

• Source run: SmolVLA_pick_object_20260408_115645
• Exported checkpoint: 050000
• Policy type: smolvla
• Dataset id: pick_object_local

Features

• Input features: observation.images.cam_head, observation.images.cam_wrist, observation.state
• Output features: action

Files

• model.safetensors: policy weights
• config.json: policy config
• train_config.json: training config snapshot
• policy_preprocessor*.json|safetensors: input processing pipeline
• policy_postprocessor*.json|safetensors: output processing pipeline

Inference

python examples/learning_il/01a_demo_ACT.py \
  --task PushT \
  --hf_repo z18820636149/SmolVLA_pick_and_place_v2 \
  --inference

Adjust --task to match the environment you evaluate against.

Notes

Local-only paths in train_config.json were sanitized during export unless --keep-local-paths was used.