SAC Go2 walking policy - 1.74M steps, 21m forward distance

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+go2_walking.mp4 filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,104 @@
+---
+tags:
+  - reinforcement-learning
+  - robotics
+  - mujoco
+  - locomotion
+  - unitree
+  - go2
+  - quadruped
+  - sac
+  - stable-baselines3
+  - strands-robots
+library_name: stable-baselines3
+model-index:
+  - name: SAC-Unitree-Go2-MuJoCo
+    results:
+      - task:
+          type: reinforcement-learning
+          name: Quadruped Locomotion
+        dataset:
+          type: custom
+          name: MuJoCo LocomotionEnv
+        metrics:
+          - type: mean_reward
+            value: 4912
+            name: Best Mean Reward
+          - type: mean_distance
+            value: 21.0
+            name: Mean Forward Distance (m)
+---
+
+# SAC Unitree Go2 — MuJoCo Locomotion Policy
+
+A **Soft Actor-Critic (SAC)** policy trained to make the Unitree Go2 quadruped **walk forward** in MuJoCo simulation.
+
+Trained entirely on a MacBook (CPU, no GPU, no Isaac Gym) using [strands-robots](https://github.com/cagataycali/strands-gtc-nvidia).
+
+## Results
+
+| Metric | Value |
+|--------|-------|
+| Algorithm | SAC (Soft Actor-Critic) |
+| Training steps | 1.74M |
+| Training time | ~40 min (MacBook M-series, CPU) |
+| Parallel envs | 8 |
+| Network | MLP [256, 256] |
+| Best reward | **4,912** |
+| Mean distance | **21 meters** per episode |
+| Forward velocity | ~1 m/s |
+| Episode length | 1,000/1,000 (full episodes) |
+
+## Demo Video
+
+See `go2_walking.mp4` — the Go2 executing the trained policy in MuJoCo.
+
+## Usage
+
+```python
+from stable_baselines3 import SAC
+
+model = SAC.load("best/best_model")
+
+# In a MuJoCo Go2 environment:
+obs, _ = env.reset()
+for _ in range(1000):
+    action, _ = model.predict(obs, deterministic=True)
+    obs, reward, done, truncated, info = env.step(action)
+```
+
+## Reward Function
+
+```
+reward = forward_vel × 5.0       # primary: move forward
+       + alive_bonus × 1.0       # stay upright
+       + upright_reward × 0.3    # orientation bonus
+       - ctrl_cost × 0.001       # minimize energy
+       - lateral_penalty × 0.3   # don't drift sideways
+       - smoothness × 0.0001     # discourage jerky motion
+```
+
+## Why SAC > PPO
+
+PPO (500K steps): Go2 learned to stand still. Reward = 615, distance = 0.02m.
+SAC (1.74M steps): Go2 walks 21 meters. Reward = 4,912.
+
+SAC's off-policy learning + entropy regularization explores more effectively in continuous action spaces.
+
+## Files
+
+- `best/best_model.zip` — Best checkpoint (highest eval reward)
+- `checkpoints/` — All 100K-step checkpoints
+- `logs/evaluations.npz` — Evaluation metrics over training
+- `go2_walking.mp4` — Demo video
+
+## Environment
+
+- **Simulator**: MuJoCo (via mujoco-python)
+- **Robot**: Unitree Go2 (12 DOF) from MuJoCo Menagerie
+- **Observation**: joint positions, velocities, torso orientation, height (37-dim)
+- **Action**: joint torques (12-dim, continuous)
+
+## License
+
+Apache-2.0

best/best_model.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1be75beaf0169882394e51b409514411622b4925458500fe3f613580bbbf316
+size 3518124

checkpoints/sac_go2_1000000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9e34811e96154ea9fc359084a8a229ce41f5499173337b069fbaeaab515fa21
+size 3518124

checkpoints/sac_go2_100000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:089bd3806fab7303f7f47451cda1eb6b94551b86c248dd3ad331a6512a710d9c
+size 3518121

checkpoints/sac_go2_1100000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa67b6823a90ce4ef15d0daa931e4ad5aba0ee55898cc44cae74c49c4007eaaa
+size 3518124

checkpoints/sac_go2_1200000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2486b8cdc3cc85dd5d8a3c8f964761ffc85e2a2a822c7fb8da7b1ec309f3cb1e
+size 3518124

checkpoints/sac_go2_1300000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc296a4819b8402a9a32f209e0ff90646ece9a51d0e3c57a26710b50e9b2ab00
+size 3518124

checkpoints/sac_go2_1400000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cd04d348dccfc8bcf6e48a54265acba01da7849dcc03c606b4fa536a0ba6096
+size 3518124

checkpoints/sac_go2_1500000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac25dffb29d95188e380e8ac0db7d36a4099b3096947a9323a0bebdaeef35fcf
+size 3518124

checkpoints/sac_go2_1600000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d862ab7c45da950588b8c9c6807015f3cc56c0c3e12f7b8170d84f605e733b04
+size 3518124

checkpoints/sac_go2_1700000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df99a844639417e7c960865286f4846b95aa3aae7b10b858a63e30852e56fc7b
+size 3518124

checkpoints/sac_go2_200000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f32315804e25be4c0f43ac746e496cac383c91abec766fd49835dcef9e500538
+size 3518121

checkpoints/sac_go2_300000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6498240d98962c000e043b6cf5351110247263ff7c3ddeb2b87be0ad80d7db4c
+size 3518122

checkpoints/sac_go2_400000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ca61bd3752cb95e0926bb70e3575031c382f4d0a74dd51e0d668d0259e5c7b3
+size 3518122

checkpoints/sac_go2_500000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9c4a43c721cb3487fd2f4ce5250f426f683fea43109f03895c59dc33da1d6fc
+size 3518122

checkpoints/sac_go2_600000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7019796c129420228764ea85522534caacded8c70c87e856e70387cce0309ec
+size 3518131

checkpoints/sac_go2_700000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:752eb6fc89aa892e797337e13f6d1ec861a014523f9573e4e5c6566b34ec4c71
+size 3518122

checkpoints/sac_go2_800000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afefe902f24986e5a34544b0933be81ffd946c230f6d1a914c42f6d18255c000
+size 3518122

checkpoints/sac_go2_900000_steps.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb0a32f7547fbeaceda54e51b673d0ee278cf0ce06467f320eb2f4a23b497227
+size 3518123

go2_walking.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:deeb18b0caf206543ecb0e9cec44633526143a5cf1bd744fad520ab65fcf759b
+size 1119500

logs/evaluations.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c13ff82c7c7544bdfe544741052312a234e2b0a410a10199dd4a56dd29699573
+size 16082