Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,5 @@ 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
+assets/drq-g1-walk.mp4 filter=lfs diff=lfs merge=lfs -text
+assets/drq-h1-walk.mp4 filter=lfs diff=lfs merge=lfs -text

DRQ+HBench-g1-walk-v0+0/agent_var.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ffa719406b465edac92cba54cec5fe9cedbd1156372351276b5ae214023c97f
+size 1086

DRQ+HBench-g1-walk-v0+0/encoder.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d5709d80a41804bc2022597b7f9bb059012adfe44336b99b5e98747f00aa9f7
+size 11368742

DRQ+HBench-g1-walk-v0+0/encoder_optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:182b7e07cd6e955732580993426602ae3846868e171da2cd4164bf71d7a478f9
+size 22741646

DRQ+HBench-g1-walk-v0+0/encoder_target.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4519493bed30b9674f828e0d2e6eebdde882e1fccb7692181e954af6da81106a
+size 11369202

DRQ+HBench-g1-walk-v0+0/exp_var.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c58e80359a1d6278882ee6ff609f2671549abd272582814c4a586e98825f6090
+size 8397

DRQ+HBench-g1-walk-v0+0/policy.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23cad46e4f7abf716f19f6002aa945dfd287317625e753171f96d7229132a62a
+size 2150846

DRQ+HBench-g1-walk-v0+0/policy_optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7745a0565e61da7361115a35b77b4a1a99b8fc4e6f393fcea1f0fa2570be621c
+size 4302822

DRQ+HBench-g1-walk-v0+0/policy_target.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbfb5e81866ecba5cea29bb5360f31232f201987c39386357a32a7042f9ada58
+size 2151236

DRQ+HBench-g1-walk-v0+0/value.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dec16fa73d702cc4b52502e78147fe02b3bb4ab2dd4dd81e0510f16643f28286
+size 6312510

DRQ+HBench-g1-walk-v0+0/value_optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34b188af7079e84c4ece58da2b53369217996489d7b440e13165695219f8f2b2
+size 12629798

DRQ+HBench-g1-walk-v0+0/value_target.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bba0c59969f24a77b4edad2581b6bfce0e494d81de76b106f36c343b36b3714a
+size 6313610

DRQ+HBench-h1-walk-v0+0/agent_var.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebf8d0bd97de9b41f974c43021363aad7db2381ebcb6b6c979e9765157850562
+size 1086

DRQ+HBench-h1-walk-v0+0/encoder.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e97535acf118e12c157146c5f1920a5c05e134d5b1297f97849e21e0d7f393f
+size 11256614

DRQ+HBench-h1-walk-v0+0/encoder_optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0dda7c1360446938d1ac6814d8a04cf816b4e54b653f42d23f9a7b5cdc4ad95
+size 22517390

DRQ+HBench-h1-walk-v0+0/encoder_target.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d4eb7b1a30f7aeb54d19c472de990125fa462b0b5a963e29aa455c4f1ea32b6
+size 11257074

DRQ+HBench-h1-walk-v0+0/exp_var.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcef3390a2ac13d6dcefa27a154272625f6c9fe3bd2f19fcb74a2cd2a2eda39f
+size 8397

DRQ+HBench-h1-walk-v0+0/policy.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3452a05d052112002baef80a53f0bc9df9e15319ce5d3a299daedeb1025b284c
+size 2142654

DRQ+HBench-h1-walk-v0+0/policy_optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9e5e806221377df42d0582874ea303b8040663c3475ef732eefc486ff607865
+size 4286438

DRQ+HBench-h1-walk-v0+0/policy_target.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34641cf8117f02057986699223530742356f5cbc40ef3aacce3ec2fec0defaed
+size 2143044

DRQ+HBench-h1-walk-v0+0/value.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:785533e0635a67da696a29613e583834435d871a72dc7a71875f630235205564
+size 6312510

DRQ+HBench-h1-walk-v0+0/value_optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48a7090fe731329eb703a4ec340b61626acb337d803a5ac751db8b7b6710a4ab
+size 12629798

DRQ+HBench-h1-walk-v0+0/value_target.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2604d82eac11d136c8480cb56960bb07097df5e63a64233c13bb6450c36cf550
+size 6313610

README.md

@@ -0,0 +1,167 @@
+---
+library_name: drq
+tags:
+  - reinforcement-learning
+  - humanoid
+  - mujoco
+  - humanoid-bench
+  - locomotion
+  - unitree-h1
+  - unitree-g1
+datasets:
+  - carlosferrazza/humanoid-bench
+base_model: dmux/DR.Q
+license: mit
+---
+
+# HumanoidBench-DR.Q · 自训通关 checkpoints
+
+_Self-trained DR.Q checkpoints that **beat** the public dmux/DR.Q baseline on HumanoidBench locomotion tasks._
+
+DR.Q 是 TD3 + model-based 表征学习的离策略 RL 算法（encoder + policy ~13 MB 推理）。
+本仓库收录在 [HumanoidBench](https://github.com/carlosferrazza/humanoid-bench) 上**从零自训通关**的 checkpoints。
+
+_DR.Q is a TD3-family off-policy RL algorithm with model-based representation learning (~13 MB encoder + policy at inference).
+This repo hosts checkpoints **trained from scratch** that pass the HumanoidBench locomotion bar._
+
+---
+
+## 📊 性能 / Performance
+
+| Task | success_rate | mean_return | N | 对比公开 baseline |
+|---|---|---|---|---|
+| **`h1-walk-v0`**  | **90%** | **801.05** | 10 ep × seed 0 | dmux/DR.Q seed 0: ~30% / mean ~530 |
+| **`g1-walk-v0`**  | **70%** | **710.52** | 10 ep × seed 0 | torque baseline: 0% / mean ~100 (**7.1× 提升**) |
+
+`success_bar = 700` (HumanoidBench locomotion threshold).
+_All numbers from deterministic eval with `action_repeat=2`. Raw JSONL in `eval/`._
+
+---
+
+## 🎬 演示 / Demos
+
+### H1-walk-v0 (Unitree H1, 19 DoF)
+
+<video controls width="720" src="https://huggingface.co/wsagi/HumanoidBench-DR.Q/resolve/main/assets/drq-h1-walk.mp4"></video>
+
+### G1-walk-v0 (Unitree G1, 23 DoF with BlockedHands wrapper)
+
+<video controls width="720" src="https://huggingface.co/wsagi/HumanoidBench-DR.Q/resolve/main/assets/drq-g1-walk.mp4"></video>
+
+---
+
+## 📦 仓库内容 / Repo layout
+
+```
+HumanoidBench-DR.Q/
+├── DRQ+HBench-h1-walk-v0+0/        # H1-walk self-trained ckpt (76 MB)
+│   ├── encoder.pt   policy.pt   agent_var.npy        ← inference (~13 MB)
+│   ├── *_target.pt × 3                                ← Q-learning targets
+│   ├── *_optimizer.pt × 3                             ← Adam states (resume)
+│   ├── value.pt                                       ← critic
+│   └── exp_var.npy                                    ← exploration variance
+├── DRQ+HBench-g1-walk-v0+0/        # G1-walk self-trained ckpt (76 MB)
+│   └── ... (same 11-file layout)
+├── eval/                            # Final eval JSONL (per-episode + summary row)
+└── assets/                          # MP4 demos
+```
+
+**推理只需 3 个文件**：`encoder.pt` + `policy.pt` + `agent_var.npy`（共 ~13 MB）。
+其余 8 个文件用于续训与 Q-learning target。
+
+_Only 3 files are needed for inference; the rest are for resume-training and Q-learning targets._
+
+---
+
+## 🚀 加载与推理 / Load & inference
+
+```python
+# Minimal inference loader — see scripts/drq_viewer.py in companion repo
+import torch, numpy as np
+from huggingface_hub import snapshot_download
+
+ckpt_dir = snapshot_download(
+    repo_id="wsagi/HumanoidBench-DR.Q",
+    allow_patterns="DRQ+HBench-h1-walk-v0+0/*",
+)
+
+var = np.load(f"{ckpt_dir}/DRQ+HBench-h1-walk-v0+0/agent_var.npy", allow_pickle=True).item()
+# encoder = Encoder(obs_dim, ...);  encoder.load_state_dict(torch.load(.../encoder.pt))
+# policy  = Policy(...);            policy.load_state_dict(torch.load(.../policy.pt))
+```
+
+完整加载链路参考配套仓库 [vitorcen/humanoid-training](https://github.com/vitorcen/humanoid-training) 的 `scripts/drq_viewer.py`。
+
+```bash
+git clone --recursive https://github.com/vitorcen/humanoid-training
+cd humanoid-training
+bash patches/apply.sh                     # apply DR.Q + HumanoidBench local patches
+DISPLAY=:0 python scripts/drq_viewer.py --task h1-walk-v0 --seed 0
+DISPLAY=:0 python scripts/drq_viewer.py --task g1-walk-v0 --seed 0
+```
+
+---
+
+## ⚠️ G1-walk 必备 patches / Required patches for G1
+
+G1 通关**不是开箱即用**，需两层 patch（详见 [g1_training_strategies.html](https://github.com/vitorcen/humanoid-training/blob/main/docs/g1_training_strategies.html)）：
+
+| Patch | 作用 |
+|---|---|
+| `patches/g1-pos-control.patch` | G1 默认 torque control → **PD position control**（与 H1 一致），sample efficiency 4×↑ |
+| `patches/humanoid-bench-g1-blocked-hands.patch` | 扩展 `BlockedHandsLocoWrapper` 支持 G1，**屏蔽 14 维手指**（37D → 23D action），避免噪声污染 encoder dynamics loss |
+
+_G1 raw torque baseline trained 1M steps and stayed at 0% / mean ~100. The combined patches lift it to 70% / mean 711._
+
+**根因（OpenCode deepseek-v4-pro 诊断）**：DR.Q 同方差 σ=0.2 exploration noise 在 37D action 上几乎每一步都扰动手指，encoder 的 dynamics loss 被迫学习无关的手指动力学 → 250k step 时 catastrophic forgetting。
+
+_Root cause: isotropic σ=0.2 noise contaminates the encoder's dynamics loss with irrelevant 14-DoF finger motion, leading to catastrophic forgetting around 250k steps._
+
+---
+
+## 🔧 训练配置 / Training config
+
+| | H1-walk | G1-walk |
+|---|---|---|
+| Algorithm | DR.Q (TD3 + zs encoder) | DR.Q + PD control + BlockedHands |
+| Env steps | 500,000 | 500,000 |
+| Wall time | 6.6 h | 3.0 h |
+| GPU | RTX 4090 | RTX 4090 |
+| `action_repeat` | 2 | 2 |
+| `save_freq` | 50,000 | 50,000 |
+| Watcher | slice-based auto-eval + early-stop (LeIsaac-inspired) | same |
+
+**训练流水线**（三个并行进程，详见配套仓库 README）：
+- A) DR.Q `main.py` 主训
+- B) `scripts/train_watcher.py` — 分 10 slice 实时聚合 + PROGRESS/UNDERFIT/OVERFIT/DEAD 早停
+- C) `scripts/ckpt_eval_loop.py` — 每出 ckpt 自动 mirror 到 HF cache + N=3 deterministic eval
+
+---
+
+## 📚 引用 / Citations
+
+```bibtex
+@article{sferrazza2024humanoidbench,
+  title={HumanoidBench: Simulated Humanoid Benchmark for Whole-Body Locomotion and Manipulation},
+  author={Sferrazza, Carmelo and Huang, Dun-Ming and Lin, Xingyu and Lee, Youngwoon and Abbeel, Pieter},
+  journal={Robotics: Science and Systems},
+  year={2024}
+}
+
+@article{yarats2022mastering,
+  title={Mastering Visual Continuous Control: Improved Data-Augmented Reinforcement Learning},
+  author={Yarats, Denis and Fergus, Rob and Lazaric, Alessandro and Pinto, Lerrel},
+  journal={ICLR},
+  year={2022}
+}
+```
+
+---
+
+## 📄 License
+
+MIT — same as base DR.Q and HumanoidBench.
+
+---
+
+_Companion repository_: [github.com/vitorcen/humanoid-training](https://github.com/vitorcen/humanoid-training) — full training scripts, patches, eval harness, and analysis docs.

assets/drq-g1-walk.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30330124bcd101284206c6abf20738b88fcb4a24d79defbf1b3e8a6138a72fc8
+size 411125

assets/drq-h1-walk.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e5b7f0a44d93ad43cf10ac0f6b6be5eb62470568d74a7f581e0c80813a5c943
+size 276249

eval/g1-walk-v0.jsonl

@@ -0,0 +1,11 @@
+{"task": "g1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 0, "ep_steps": 1000, "ep_return": 750.7662936729176, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.8296611309051514}
+{"task": "g1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 1, "ep_steps": 1000, "ep_return": 733.3856304355644, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.8071818351745605}
+{"task": "g1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 2, "ep_steps": 1000, "ep_return": 521.2008137157275, "success": false, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.8442943096160889}
+{"task": "g1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 3, "ep_steps": 1000, "ep_return": 706.1093643279406, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.7915904521942139}
+{"task": "g1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 4, "ep_steps": 1000, "ep_return": 657.5082534046979, "success": false, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.8038837909698486}
+{"task": "g1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 5, "ep_steps": 1000, "ep_return": 651.3662470092779, "success": false, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.8597824573516846}
+{"task": "g1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 6, "ep_steps": 1000, "ep_return": 769.2891580176396, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.7870059013366699}
+{"task": "g1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 7, "ep_steps": 1000, "ep_return": 814.1963970582311, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.8006560802459717}
+{"task": "g1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 8, "ep_steps": 1000, "ep_return": 735.2333430160378, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.8080053329467773}
+{"task": "g1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 9, "ep_steps": 1000, "ep_return": 766.1107264076938, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.807042121887207}
+{"_summary": true, "n_episodes": 10, "success_rate": 0.7, "mean_return": 710.5166227065728, "mean_steps": 1000.0, "mean_subtasks": 0.0, "mean_ttf_success_s": null, "timeout_rate": 1.0, "task": "g1-walk-v0", "driver": "drq", "seeds": [0], "eval_per_seed": 10, "total_wall_s": 8.86957573890686}

eval/h1-walk-v0.jsonl

@@ -0,0 +1,11 @@
+{"task": "h1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 0, "ep_steps": 1000, "ep_return": 831.7330195042176, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.4069836139678955}
+{"task": "h1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 1, "ep_steps": 1000, "ep_return": 830.7659604815722, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.39011359214782715}
+{"task": "h1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 2, "ep_steps": 1000, "ep_return": 831.6017931113811, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.3844790458679199}
+{"task": "h1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 3, "ep_steps": 1000, "ep_return": 830.9254211628348, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.38822317123413086}
+{"task": "h1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 4, "ep_steps": 1000, "ep_return": 831.2334128876776, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.38614797592163086}
+{"task": "h1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 5, "ep_steps": 753, "ep_return": 530.3864998710653, "success": false, "subtasks_max": 0, "time_to_success_s": null, "timed_out": false, "wall_time_s": 0.3132052421569824}
+{"task": "h1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 6, "ep_steps": 1000, "ep_return": 830.451588613047, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.3938441276550293}
+{"task": "h1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 7, "ep_steps": 1000, "ep_return": 830.9401017369254, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.3912630081176758}
+{"task": "h1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 8, "ep_steps": 1000, "ep_return": 830.5614255749334, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.4045600891113281}
+{"task": "h1-walk-v0", "driver": "drq", "seed": 0, "ep_idx": 9, "ep_steps": 1000, "ep_return": 831.8926512136476, "success": true, "subtasks_max": 0, "time_to_success_s": null, "timed_out": true, "wall_time_s": 0.38388848304748535}
+{"_summary": true, "n_episodes": 10, "success_rate": 0.9, "mean_return": 801.0491874157302, "mean_steps": 975.3, "mean_subtasks": 0.0, "mean_ttf_success_s": null, "timeout_rate": 0.9, "task": "h1-walk-v0", "driver": "drq", "seeds": [0], "eval_per_seed": 10, "total_wall_s": 4.451604843139648}