Instructions to use drashutoshspace/moonbot_act_no_tf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LeRobot
How to use drashutoshspace/moonbot_act_no_tf with LeRobot:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
moonbot ACT β three blocks stack, WITHOUT force/torque
ACT trained from scratch (ImageNet ResNet-18 backbone). LeRobot ACT does not take the language instruction as input, so it is a no-language baseline across the 9 tasks, trained on gdiazsrl/lerobot_sep23_no_tf (296 episodes, 519,684 frames at 30 Hz, 9 language instructions; MoonBot 7-DOF arm, block stacking).
Input/output: 15-dim state = joint_read (8: 7 joints + gripper) + tip_pos (7: end-effector pose), no F_ee force/torque; 3 RGB cameras (front, eef, right); 8-dim action (7 joint positions + gripper).
Comparison run. One of six policies comparing pi0 / SmolVLA / ACT with vs without force/torque, all trained with batch size 16, 20,000 steps, seed 1000 and no validation split; otherwise each uses its own LeRobot preset. The two datasets are the same episodes and frames, except that the 6 F_ee dimensions are removed from the state in the no-F/T one. The counterpart of this model with force/torque is moonbot_act_tf.
| model | F/T | repo | W&B run |
|---|---|---|---|
| pi0 | with | moonbot_pi0_three_blocks_stack |
bkbszs91 |
| pi0 | without | moonbot_pi0_no_tf |
ahb0i6l9 |
| SmolVLA | with | moonbot_smolvla_tf |
rn1je6y9 |
| SmolVLA | without | moonbot_smolvla_no_tf |
9i3olvax |
| ACT | with | moonbot_act_tf |
uw4alfzk |
| ACT | without | moonbot_act_no_tf |
a7hpuoto |
Training
| initialization | from scratch, ResNet-18 backbone with ImageNet weights |
| parameters | 52M (all trainable) |
| language input | no |
| action chunk | 100 (all 100 executed) |
| optimizer | AdamW, lr 1e-5, betas (0.9, 0.999), weight decay 1e-4, grad clip 10 |
| LR schedule | constant |
| batch size / steps | 16 / 20,000 (about 0.62 epochs) |
| seed | 1000 |
| GPU | 1x NVIDIA RTX PRO 6000 Blackwell (96 GB) |
| training time | 1.4 h (GPU shared with another run) |
| final training loss | 0.085 (this policy's own objective; not comparable across algorithms) |
| framework | LeRobot 0.6.1 |
| training curves | https://wandb.ai/drmishra-space/lerobot/runs/a7hpuoto |
Checkpoints
checkpoint-005000, checkpoint-010000, checkpoint-015000, checkpoint-020000 (final). Each checkpoint-XXXXXX/ folder holds the weights, config.json, train_config.json, the pre/post-processor files, this run's rosetta_contract.yaml and DEPLOY.md. checkpoint-020000/ also includes training_state/ for resuming. deploy/ holds the contract once.
Complete checkpoints, including the optimizer state, are also on the team OneDrive under DATA/Processed/act_no_tf.
Deployment
Follow DEPLOY.md in the checkpoint folder and use this folder's rosetta_contract.yaml.
Not yet offline-ready: on first load ACT downloads the ImageNet ResNet-18 weights through torchvision (they are then overwritten by the trained weights). Give the robot internet once, pre-fill TORCH_HOME, or set pretrained_backbone_weights to null in config.json.