publish backroll: perpetual, Mjlab-GotATalent-MicroDuck

README.md

@@ -0,0 +1,47 @@
+---
+tags:
+- microduck
+- robotics
+- reinforcement-learning
+- onnx
+library_name: microduck
+pipeline_tag: robotics
+---
+
+# backroll
+
+Mjlab-GotATalent-MicroDuck
+
+A **perpetual** policy for the [microduck](https://github.com/pollen-robotics/microduck) (61-D observation, 14 actions, 50 Hz). Runs until told otherwise — a gait, loaded into a policy slot.
+
+## Run it on a robot
+
+```bash
+sudo robotctl policy load <slot> dashi/microduck-backroll
+```
+
+The observation normalizer is baked into `policy.onnx`; feed raw observations.
+`manifest.json` follows schema 2 of the microduck policy manifest (`docs/policy-manifest.md` in the daemon repo).
+
+## Training
+
+- **task_id**: `Mjlab-GotATalent-MicroDuck`
+- **repo**: `https://github.com/bxshi/microduck-challenges.git`
+- **branch**: `HEAD`
+- **commit**: `65971f273`
+- **checkpoint**: `750`
+- **seed**: `1`
+- **base**: `mjlab-microduck 0.1.0 @ 8d0dbc94c`
+- **started**: `2026-10-22T00:46:36Z`
+
+## Reproduce
+
+Same code, same `uv.lock`, same command, same seed. Training it again yields a comparable policy, not the same weights: GPU reinforcement learning is not bit-reproducible across machines.
+
+```bash
+git clone https://github.com/bxshi/microduck-challenges.git
+cd microduck-challenges
+git checkout 65971f273
+uv sync
+uv run train Mjlab-GotATalent-MicroDuck --env.scene.num-envs 4096 --agent.max-iterations 3000 --agent.seed 1 --agent.logger tensorboard --agent.run-name roll4
+```

checkpoint.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d4eab17de4ee03eff19e88fa643482fe7b5bfc6c0a20763cce3bfd78c24957b
+size 4849507

manifest.json

@@ -0,0 +1,58 @@
+{
+  "schema_version": 2,
+  "model_api": 1,
+  "obs_len": 61,
+  "action_len": 14,
+  "robot": {
+    "model": "microduck",
+    "hw_rev": 1,
+    "servos": "xl330",
+    "control_hz": 50,
+    "accessories": []
+  },
+  "name": "backroll",
+  "kind": "perpetual",
+  "entry_pose": "standing",
+  "description": "Mjlab-GotATalent-MicroDuck",
+  "command": {
+    "encoding": "constant",
+    "idle": [
+      0.0,
+      0.0,
+      0.0
+    ],
+    "twist": "[sit flag, held 0 (stand); roll trigger: a short pulse of 1 starts one backward roll and stand-up; unused]",
+    "head": "[neck_pitch, head_pitch, head_yaw, head_roll]: rad from the default pose",
+    "body": "unused (zeros)"
+  },
+  "duration_s": null,
+  "training": {
+    "repo": "https://github.com/bxshi/microduck-challenges.git",
+    "commit": "65971f273",
+    "branch": "HEAD",
+    "dirty": false,
+    "command": [
+      "train",
+      "Mjlab-GotATalent-MicroDuck",
+      "--env.scene.num-envs",
+      "4096",
+      "--agent.max-iterations",
+      "3000",
+      "--agent.seed",
+      "1",
+      "--agent.logger",
+      "tensorboard",
+      "--agent.run-name",
+      "roll4"
+    ],
+    "seed": 1,
+    "base": "mjlab-microduck 0.1.0 @ 8d0dbc94c",
+    "started": "2026-10-22T00:46:36Z",
+    "task_id": "Mjlab-GotATalent-MicroDuck",
+    "checkpoint": 750,
+    "source_file": "model_750.pt"
+  },
+  "arena": {
+    "event": "got-a-talent"
+  }
+}

policy.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e8ce68291e295faa20a5f1c4af88744dd629788f8370b3f78d04c252b54f809
+size 793754

timeline.json

@@ -0,0 +1,19 @@
+{
+  "timeline_version": 1,
+  "duration_s": 20.0,
+  "keyframes": [
+    {"t": 0.0,  "twist": [0, 0, 0], "head": [0, 0, 0, 0]},
+    {"t": 2.0,  "twist": [0, 1, 0], "head": [0, 0, 0, 0]},
+    {"t": 2.6,  "twist": [0, 0, 0], "head": [0, 0, 0, 0]},
+    {"t": 5.0,  "twist": [0, 1, 0], "head": [0, 0, 0, 0]},
+    {"t": 5.6,  "twist": [0, 0, 0], "head": [0, 0, 0, 0]},
+    {"t": 8.0,  "twist": [0, 1, 0], "head": [0, 0, 0, 0]},
+    {"t": 8.6,  "twist": [0, 0, 0], "head": [0, 0, 0, 0]},
+    {"t": 11.0, "twist": [0, 1, 0], "head": [0, 0, 0, 0]},
+    {"t": 11.6, "twist": [0, 0, 0], "head": [0, 0, 0, 0]},
+    {"t": 14.0, "twist": [0, 1, 0], "head": [0, 0, 0, 0]},
+    {"t": 14.6, "twist": [0, 0, 0], "head": [0, 0, 0, 0]},
+    {"t": 17.0, "twist": [0, 1, 0], "head": [0, 0, 0, 0]},
+    {"t": 17.6, "twist": [0, 0, 0], "head": [0, 0, 0, 0]}
+  ]
+}