Add MIKASA StarVLA dataset mixture
Browse files
README.md
CHANGED
|
@@ -8,7 +8,7 @@ Accepted zero-latency teacher rollouts for the supported benchmark tasks.
|
|
| 8 |
## Layout
|
| 9 |
|
| 10 |
- `zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-v1/raw`: 275 successful source NPZ episodes.
|
| 11 |
-
- `zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-v1/lerobot` and `
|
| 12 |
- `zero-latency/humanoidbench-balance-simple/teacher-rollouts-v1/lerobot` and `lerobot__val`: fixed 90/10 official GR00T split, including videos and metadata.
|
| 13 |
|
| 14 |
Extreme Parkour and HAIC datasets are intentionally not published until their current VLA pipelines pass their documented gates.
|
|
|
|
| 8 |
## Layout
|
| 9 |
|
| 10 |
- `zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-v1/raw`: 275 successful source NPZ episodes.
|
| 11 |
+
- `zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-v1/lerobot`, `lerobot__val`, and `mikasa_intercept_grab_fast_h1.json`: fixed 250/25 StarVLA split and mixture.
|
| 12 |
- `zero-latency/humanoidbench-balance-simple/teacher-rollouts-v1/lerobot` and `lerobot__val`: fixed 90/10 official GR00T split, including videos and metadata.
|
| 13 |
|
| 14 |
Extreme Parkour and HAIC datasets are intentionally not published until their current VLA pipelines pass their documented gates.
|
zero-latency/mikasa-intercept-grab-fast/teacher-rollouts-v1/mikasa_intercept_grab_fast_h1.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"local/intercept_grab_fast_vla_v0_h1_train": [
|
| 3 |
+
[
|
| 4 |
+
"local/intercept_grab_fast_vla_v0_h1_train",
|
| 5 |
+
1.0,
|
| 6 |
+
"mikasa_franka_h1"
|
| 7 |
+
]
|
| 8 |
+
],
|
| 9 |
+
"local/intercept_grab_fast_vla_v0_h1_val": [
|
| 10 |
+
[
|
| 11 |
+
"local/intercept_grab_fast_vla_v0_h1_val",
|
| 12 |
+
1.0,
|
| 13 |
+
"mikasa_franka_h1"
|
| 14 |
+
]
|
| 15 |
+
]
|
| 16 |
+
}
|