Update model card (task / checkpoint table)
Browse files
README.md
CHANGED
|
@@ -6,30 +6,50 @@ tags:
|
|
| 6 |
---
|
| 7 |
# GR00T N1.7 β Humanoid Benchmark
|
| 8 |
|
| 9 |
-
GR00T N1.7 checkpoints fine-tuned on Humanoid Benchmark tasks (
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
|
| 16 |
-
|
|
| 17 |
-
| `
|
| 18 |
-
| `
|
| 19 |
-
| `
|
| 20 |
-
| `
|
| 21 |
-
| `
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
---
|
| 7 |
# GR00T N1.7 β Humanoid Benchmark
|
| 8 |
|
| 9 |
+
GR00T N1.7 (Gr00tN1d7) checkpoints fine-tuned on Humanoid Benchmark tasks (action head fine-tuned, VLM backbone frozen). The VLM backbone `nvidia/Cosmos-Reason2-2B` is loaded from its own (gated) repository.
|
| 10 |
+
|
| 11 |
+
## Checkpoints
|
| 12 |
+
|
| 13 |
+
100 demonstrations per task. Every row is one training run; the epoch folders are its published checkpoints.
|
| 14 |
+
|
| 15 |
+
| Folder | Task | Robot | Cameras (model input) | Global batch | Proprio augmentation | Checkpoints |
|
| 16 |
+
|---|---|---|---|---|---|---|
|
| 17 |
+
| `b6-pedal-bin/` | B6 β pick up a chip barrel, step on the pedal bin, drop it in | Unitree G1 (Inspire hands) | ego_view, chest | 696 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 18 |
+
| `a3-fridge-open/` | A3 β open the fridge door | Unitree G1 (Inspire hands) | ego_view, chest | 696 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 19 |
+
| `a3-fridge-open/b256/` | A3 β open the fridge door (re-train, current protocol) | Unitree G1 (Inspire hands) | ego_view, chest | 256 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 20 |
+
| `a7-cabinet-close/` | A7 β close the cabinet door | Unitree G1 (Inspire hands) | ego_view, chest | 696 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 21 |
+
| `a7-cabinet-close/b256/` | A7 β close the cabinet door (re-train, current protocol) | Unitree G1 (Inspire hands) | ego_view, chest | 256 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 22 |
+
| `a6-door-pull/` | A6 β pull a door open | Unitree G1 (Inspire hands) | ego_view, chest | 256 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 23 |
+
| `a10-trolley-push/` | A10 β push a trolley | Unitree G1 (Inspire hands) | ego_view, chest | 696 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 24 |
+
| `a20-basket-transfer/` | A20 β transfer a basket | Unitree G1 (Inspire hands) | ego_view, chest | 696 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 25 |
+
| `b7-black-clothes-hamper/` | B7 β take the black garment off the hanger and drop it into the hamper | Unitree G1 (Inspire hands) | ego_view, chest | 696 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 26 |
+
| `c6-bathroom-door-fetch/` | C6 β open the bathroom door, take the tea bottle from the vanity, put it on the dining table | Unitree G1 (Inspire hands) | ego_view, chest | 256 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 27 |
+
| `a18-pour-water/` | A18 β pour water from a pitcher into a glass | H2 (Sharpa hands) | head, chest | 696 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 28 |
+
| `a17-unscrew-bottle/` | A17 β unscrew the cap of a bottle | H2 (Sharpa hands) | head, chest | 696 | aligned (GR00T A20 recipe) | `epoch-010`, `epoch-020` |
|
| 29 |
+
|
| 30 |
+
A `b256/` or `aligned/` sub-folder is a re-train of that task under the current benchmark protocol (2026-11-12: global
|
| 31 |
+
batch GR00T / pi0.5 / Delta Flow / ACT 256, Psi-Zero 512, DiT4DiT 64; proprioception augmentation aligned with the
|
| 32 |
+
GR00T A20 recipe) and is the one to use for that task; the checkpoints directly under the task folder are the earlier
|
| 33 |
+
training round, kept for reference.
|
| 34 |
+
|
| 35 |
+
## What an epoch folder contains
|
| 36 |
+
|
| 37 |
+
Every `epoch-XXX/` folder is self-contained for evaluation and deployment: weights, model configuration, normalization
|
| 38 |
+
statistics and `protocol.json` (task, data, input/output dimensions, cameras, batch, steps, augmentation, file list).
|
| 39 |
+
Simple training records are included for reproducibility (training plan, training log, configs, metrics); optimizer /
|
| 40 |
+
scheduler / RNG / DeepSpeed resume state is not. `code/` holds the adapter code used for training and inference
|
| 41 |
+
(`code/VERSIONS.json`: upstream commits and package versions).
|
| 42 |
+
|
| 43 |
+
## Inputs and augmentation
|
| 44 |
+
|
| 45 |
+
Model inputs are identical across all benchmark models: G1 state 44-D (29 body joint angles + 6 + 6 measured hand values
|
| 46 |
+
+ 3 projected gravity), H2 state 78-D (75 joint angles + 3 projected gravity), no joint velocities. Cameras: G1 ego view +
|
| 47 |
+
chest; H2 head + chest (A14: head + chest + wrist). The third-person camera recorded in the datasets is for data review
|
| 48 |
+
only and is never a model input. Each camera is a separate image input, except for Delta Flow and DiT4DiT, which
|
| 49 |
+
place the views side by side in one frame.
|
| 50 |
+
Training uses the Humanoid Benchmark online data augmentation spec v2 (15 photometric / geometric / sensor / lighting
|
| 51 |
+
effects, one of them boosted to 1.5x per sample) with the model's own image augmentation turned off, plus joint-angle
|
| 52 |
+
bias and noise on the proprioception (column "Proprio augmentation"; the earlier recipe also delayed the
|
| 53 |
+
proprioception and dropped action-history rows).
|
| 54 |
+
|
| 55 |
+
Earlier repository states are kept in the git history (e.g. revision `2e8f375584631cae0454911e58b5b7850ea9f678`, an older batch setting).
|