pravsels commited on
Commit
33f596e
·
verified ·
1 Parent(s): 17aa1aa

Update model card

Browse files
Files changed (1) hide show
  1. README.md +33 -0
README.md ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: gr00t
3
+ tags:
4
+ - gr00t
5
+ - gr00t-n1.7
6
+ - robotics
7
+ - vla
8
+
9
+ base_model: nvidia/GR00T-N1.7-3B
10
+ ---
11
+
12
+ # groot1.7_block_stack_20k
13
+
14
+ Fine-tuned [GR00T N1.7-3B](https://huggingface.co/nvidia/GR00T-N1.7-3B) for `block_stack` on SO101 data.
15
+
16
+ | | |
17
+ |---|---|
18
+ | **Base model** | [nvidia/GR00T-N1.7-3B](https://huggingface.co/nvidia/GR00T-N1.7-3B) |
19
+ | **Dataset** | [villekuosmanen/armnetbench_block_stack](https://huggingface.co/datasets/villekuosmanen/armnetbench_block_stack) |
20
+ | **Task** | `block_stack` |
21
+ | **Embodiment** | single-arm SO-101 |
22
+ | **Training** | 20k steps, batch 32 (8/GPU × 4 DDP), 4× RTX 5090 on Vast.ai, load_bf16=True |
23
+ | **Prior HF repo** | [pravsels/groot1.7_block_stack](https://huggingface.co/pravsels/groot1.7_block_stack) |
24
+ | **W&B project** | [groot1.7_block_stack](https://wandb.ai/pravsels/groot1.7_block_stack) |
25
+ | **W&B run** | [pwazfb2z](https://wandb.ai/pravsels/groot1.7_block_stack/runs/pwazfb2z) |
26
+
27
+ ## Checkpoints
28
+
29
+ Final checkpoint (step `20000`) lives at the **repository root** for direct loading.
30
+
31
+ ## Usage
32
+
33
+ Load with the GR00T inference stack from this repo root (weights + processor configs).