drwahl commited on
Commit
aff045f
·
verified ·
1 Parent(s): cbcf1e3

Add the model card

Browse files
Files changed (1) hide show
  1. README.md +23 -0
README.md ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: lerobot/smolvla_base
4
+ pipeline_tag: robotics
5
+ tags:
6
+ - LeRobot
7
+ - so101
8
+ - smolvla
9
+ - lora
10
+ ---
11
+
12
+ # smolvla_so101_block_stack_sim
13
+
14
+ SmolVLA fine-tuned to stack one 30 mm block on another with an [SO-101](https://github.com/TheRobotStudio/SO-ARM100), in simulation. Trained by `train/smolvla_so101.yaml` in [vla-test](https://github.com/danwahl/vla-test) on [drwahl/so101_block_stack_sim](https://huggingface.co/datasets/drwahl/so101_block_stack_sim).
15
+
16
+ LoRA (r=32, alpha 64) on the SmolVLM2 language backbone of `lerobot/smolvla_base`, with the action expert and the projection heads dense, SigLIP frozen: 20k steps at batch 4, the protocol [drwahl/pi05_so101_block_stack_sim](https://huggingface.co/drwahl/pi05_so101_block_stack_sim) was trained on. Normalization, learning rate and chunk length are the base checkpoint's own. `adapter/` holds the adapter and the processors, which carry the normalization and the camera renaming; lerobot loads it onto the base weights.
17
+
18
+ ```bash
19
+ hf download drwahl/smolvla_so101_block_stack_sim --local-dir smolvla_so101_block_stack_sim
20
+ uv run python sim/scripts/eval.py smolvla_so101_block_stack_sim/adapter --no-rtc --no-smooth
21
+ ```
22
+
23
+ The eval rolls the policy out on the 150 held-out layouts and scores it by the gates that screened the oracle. Executing its whole 50-step chunk, the setting that stacks most on one seed, it stacks 0.3 of 150 and lifts 11.3, averaged over three seeds. It predicts the demonstrations' actions on their own frames about as closely as the other policies do, and does not recover from its own errors once it drives the arm.