Publish evaluated headstand checkpoint and ONNX export

README.md

@@ -0,0 +1,24 @@
+---
+library_name: onnx
+pipeline_tag: reinforcement-learning
+tags:
+- robotics
+- mujoco
+- microduck
+---
+# Microduck headstand: backroll
+
+Simulation policy for `Mjlab-HeadstandBackrollLegsTogether-Flat-MicroDuck`. The original checkpoint and an ONNX export with the observation normalizer included are provided. Input: 61 observation values; output: 14 joint actions.
+
+The original checkpoint passed 32 of 32 individual simulated attempts at seed 0 under the documented evaluator. This does not establish hardware performance or success from arbitrary starting poses. The ONNX export is provided separately; the published rollout counts were measured with the original checkpoint.
+
+Files:
+
+- `model_1999.pt`: the unchanged evaluated checkpoint.
+- `policy.onnx`: export through the training repository's standard exporter.
+- `provenance.json`: source run, hashes, and code provenance.
+- `evaluation.json`: individual evaluation report.
+
+The six policies together completed 87 of 96 routine attempts across seeds 0, 1, and 2. See the [evaluation code, setup, and raw results](https://github.com/zachgarner/microduck-headstand/tree/71ad395). The saved checkpoint can be placed in the evaluator's cache at `logs/rsl_rl/microduck_headstand_backroll_legs_together/wandb_checkpoints/ax1vgv8z/model_1999.pt`; no W&B download is then needed for that checkpoint.
+
+These are simulation research artifacts, not a robot-daemon installation package. No runtime manifest or hardware installation command is supplied. The routine uses contact-based handovers and a separate standing policy, and has not been tested on a physical robot. Head-force samples are taken at 50 Hz and may miss shorter impacts. Exit success checks final standing, not leg shape throughout the exit.

evaluation.json

@@ -0,0 +1,150 @@
+{
+  "task": "Mjlab-HeadstandBackrollLegsTogether-Flat-MicroDuck",
+  "run": "ax1vgv8z",
+  "checkpoint": "model_1999.pt",
+  "checkpoint_sha256": "e94cd3256060f09004a3b6210545fdfbe29b5e64084df697d37677e8ae19e8fd",
+  "seed": 0,
+  "bucket": "hold",
+  "episodes": 32,
+  "steps": 298,
+  "seconds": 5.96,
+  "successes": 32,
+  "median_time_to_goal_s": 0.52,
+  "sampled_head_force_median_n": null,
+  "sampled_head_force_max_n": null,
+  "reset_count": 0,
+  "tool_sha256": {
+    "check32.py": "d70963e6c1dcc69e311e9c57041d4d3e3a236532790321bfc39b9c6f8e3d4f51",
+    "eval_checkpoint.py": "bb12b02d3a1562b350a008bf6436a3adc4be1f187f8dbaae44efc9b24a60a893"
+  },
+  "per_episode": [
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.5
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.5
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.5
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.5
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.5
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.54
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    },
+    {
+      "success": true,
+      "first_goal_s": 0.52
+    }
+  ]
+}

model_1999.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e94cd3256060f09004a3b6210545fdfbe29b5e64084df697d37677e8ae19e8fd
+size 4831157

policy.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ff4753958f6d42c9b62254b0926f917e513a7a20edd224dab3396410ef3f10d
+size 793816

provenance.json

@@ -0,0 +1,11 @@
+{
+  "task": "Mjlab-HeadstandBackrollLegsTogether-Flat-MicroDuck",
+  "wandb_run": "zachgarner-ai/mjlab_microduck/ax1vgv8z",
+  "checkpoint": "model_1999.pt",
+  "checkpoint_sha256": "e94cd3256060f09004a3b6210545fdfbe29b5e64084df697d37677e8ae19e8fd",
+  "onnx_sha256": "8ff4753958f6d42c9b62254b0926f917e513a7a20edd224dab3396410ef3f10d",
+  "evaluated_training_commit": "d0680f3d94774d612e7e90bea2eb1635e4eed869",
+  "evaluation_repository": "https://github.com/zachgarner/microduck-headstand/tree/71ad395",
+  "exporter": "mjlab_microduck.export.run_export (observation normalizer included)",
+  "scope": "Simulation artifacts; not a robot-daemon installation package."
+}