Datasets:

ccnets
/

causal-gpt-rl-datasets

@@ -1,149 +1,155 @@
----
-license: cc-by-4.0
-viewer: false
-tags:
-  - reinforcement-learning
-  - offline-reinforcement-learning
-  - mujoco
-  - minari
-  - causal-gpt-rl
----
-# Causal GPT-RL — MuJoCo trajectories
-MuJoCo trajectories packaged as [Minari](https://minari.farama.org) datasets, in
-**offline-RL** form. What makes this repository different from a collection of
-rollouts is where the trajectories come from: **they were recorded by a policy
-that had never seen expert data.**
-The first dataset is **1000 episodes of `Humanoid-v5`, recorded by a policy
-trained on `simple` and `medium` trajectories alone** — one episode at every
-retention length from 1 to 1000, each value used exactly once.
-## Why this repository exists
-The scarcity of good trajectories is the binding constraint in offline RL. There
-is no internet-scale corpus of behaviour, and the usual way out is circular —
-expert datasets need an expert policy, and an expert policy is most of the
-problem already solved.
-So the question is not only *can a model learn from the current dataset*, but
-**can it produce a better next one**:
-```
-simple                     ->  model  ->  medium
-simple + medium            ->  model  ->  better than medium
-simple + medium + that     ->  model  ->  better still
-                    ... repeated, never using expert data ...
-                                       ->  expert
-```
-Each dataset here is one turn of that loop. No expert trajectories are used at
-any point.
-## Where the loop stands
-Normalized skill on the [model card's](https://huggingface.co/ccnets/causal-gpt-rl)
-scale — a random policy is `0`, expert is `100`:
-| | dataset | mean return | normalized |
-|---|---|---:|---:|
-| input | `mujoco/humanoid/simple-v0` | 5484.6 | 63.3 |
-| input | `mujoco/humanoid/medium-v0` | 7014.5 | 81.3 |
-| **this repo** | **`mujoco/humanoid-v5/medium-v0`** | **7494.3** | **86.9** |
-| target | expert | 8605.1 | 100.0 |
-The recorded trajectories score **+479.8 return above the medium dataset the
-policy was trained on**, and **85.7 % of the 1000 episodes beat that dataset's
-mean**. The gap left to expert is 13.1 normalized points.
-## `humanoid-v5/medium-v0`
-Recorded by the
-[`ccnets/causal-gpt-rl/humanoid-v5`](https://huggingface.co/ccnets/causal-gpt-rl)
-bundle, trained on the `simple` and `medium` trajectories of that environment
-with no expert data.
-| | |
-|---|---|
-| Episodes | 1000 |
-| Transitions | 924,747 |
-| Episode length | 924.7 mean (87 min, 1000 max) |
-| Return | 7494.3 ± 1667.6 (515.4 min, 8340.7 max) |
-| Ended by termination | 156 |
-| Ended by the 1000-step limit | 844 |
-| Observation / action | `Box(348,)` / `Box(17,)`, float32 |
-### One retention per episode
-Every episode carries a distinct seed and a **different retention** —
-`kv_cache_max_len` from 1 to 1000, each value used exactly once. Retention is
-how many steps of its own past the policy keeps while acting. It is a load-time
-argument, so all 1000 episodes come from the same weights and the same trained
-context length of 32; only how much history each rollout kept differs.
-`episode_metadata.jsonl` maps every episode file to the retention that recorded
-it — episodes are numbered in the order they finished, not by retention:
-```json
-{"episode": "ep_000000", "row": 4, "transitions": 113, "terminated": true, "truncated": false, "kv_cache_max_len": 16}
-```
-That makes the dataset usable two ways: as offline-RL trajectories, and as a
-measurement of what retention does to a policy's behaviour. Within it, episodes
-recorded between 16 and 128 steps of retention average 91.1 normalized against
-85.5 for those above 500 — but with one episode per retention, seed variance is
-larger than the effect, so treat the per-retention axis as evidence to build on
-rather than a settled result.
-### Files
-```
-humanoid-v5/medium-v0/
-  data/main_data.hdf5        the Minari dataset
-  data/metadata.json
-  episode_metadata.jsonl     one line per episode: file, row, length, retention
-  collection_spec.json       what recorded it: bundle, context length, retentions
-  validation_summary.json    every number quoted above
-```
-### Load it
-```python
-import minari
-dataset = minari.load_dataset("mujoco/humanoid-v5/medium-v0")
-episode = next(dataset.iterate_episodes())
-print(episode.observations.shape, episode.rewards.sum())
-```
-Place the `humanoid-v5/medium-v0` directory under your Minari datasets path
-(`~/.minari/datasets/mujoco/`), or download the repository and point
-`MINARI_DATASETS_PATH` at it.
-The files were recorded with `torch 2.12.0`, `gymnasium 1.2.2` and
-`mujoco 3.8.1`; a different simulator release is a different measurement even
-with identical weights and seeds. The recorder and packager are
-[`collection/`](https://github.com/ccnets-team/causal-gpt-rl/tree/main/collection),
-and the cycle this repository is built on is described in
-[Improving the next dataset](https://github.com/ccnets-team/causal-gpt-rl/blob/main/collection/docs/improving-the-next-dataset.md).
-## A note on names
-`mujoco/humanoid/medium-v0` and `mujoco/humanoid-v5/medium-v0` differ by one
-segment and are not the same data. The first is the **input** — a Minari dataset
-this project trained on. The second is this repository's **output** — what the
-resulting policy recorded, scoring above the first. The name marks the rung of
-the ladder the data was produced at, not a quality claim about the file.
-## Companion repos
-- Policy bundles for eight MuJoCo environments:
-  [ccnets/causal-gpt-rl](https://huggingface.co/ccnets/causal-gpt-rl)
-- Unity ML-Agents trajectories, with a calibrated quality ladder:
-  [ccnets/causal-gpt-rl-unity-datasets](https://huggingface.co/datasets/ccnets/causal-gpt-rl-unity-datasets)
-- Code:
-  [ccnets-team/causal-gpt-rl](https://github.com/ccnets-team/causal-gpt-rl)
-**한국어 카드 / Korean edition:**
-[README.ko.md](https://huggingface.co/datasets/ccnets/causal-gpt-rl-datasets/blob/main/README.ko.md)

+---
+license: cc-by-4.0
+viewer: false
+tags:
+  - reinforcement-learning
+  - offline-reinforcement-learning
+  - mujoco
+  - minari
+  - causal-gpt-rl
+---
+# Causal GPT-RL — MuJoCo trajectories
+MuJoCo trajectories packaged as [Minari](https://minari.farama.org) datasets, in
+**offline-RL** form. What makes this repository different from a collection of
+rollouts is where the trajectories come from: **they were recorded by a policy
+that had never seen expert data.**
+The first dataset is **1000 episodes of `Humanoid-v5`, recorded by a policy
+trained on `simple` and `medium` trajectories alone** — one episode at every
+retention length from 1 to 1000, each value used exactly once.
+## Why this repository exists
+The scarcity of good trajectories is the binding constraint in offline RL. There
+is no internet-scale corpus of behaviour, and the usual way out is circular —
+expert datasets need an expert policy, and an expert policy is most of the
+problem already solved.
+So the question is not only *can a model learn from the current dataset*, but
+**can it produce a better next one**:
+```
+simple                     ->  model  ->  medium
+simple + medium            ->  model  ->  better than medium
+simple + medium + that     ->  model  ->  better still
+                    ... repeated, never using expert data ...
+                                       ->  expert
+```
+Each dataset here is one turn of that loop. No expert trajectories are used at
+any point.
+## Where the loop stands
+Normalized skill on the [model card's](https://huggingface.co/ccnets/causal-gpt-rl)
+scale — a random policy is `0`, expert is `100`:
+| | dataset | mean return | normalized |
+|---|---|---:|---:|
+| input | Minari `mujoco/humanoid/simple-v0` | 5484.6 | 63.3 |
+| input | Minari `mujoco/humanoid/medium-v0` | 7014.5 | 81.3 |
+| **this repo** | **`ccnets/humanoid/medium-v0`** | **7494.3** | **86.9** |
+| target | expert | 8605.1 | 100.0 |
+"Minari" marks the two [Farama-published](https://minari.farama.org/datasets/mujoco/humanoid/)
+datasets this project trained on. Everything else in this repository is
+recorded, not downloaded.
+The recorded trajectories score **+479.8 return above the medium dataset the
+policy was trained on**, and **85.7 % of the 1000 episodes beat that dataset's
+mean**. The gap left to expert is 13.1 normalized points.
+## `humanoid/medium-v0`
+Recorded by the
+[`ccnets/causal-gpt-rl/humanoid-v5`](https://huggingface.co/ccnets/causal-gpt-rl)
+bundle, trained on the Minari `simple-v0` and `medium-v0` trajectories of that
+environment with no expert data.
+| | |
+|---|---|
+| Episodes | 1000 |
+| Transitions | 924,747 |
+| Episode length | 924.7 mean (87 min, 1000 max) |
+| Return | 7494.3 ± 1667.6 (515.4 min, 8340.7 max) |
+| Ended by termination | 156 |
+| Ended by the 1000-step limit | 844 |
+| Observation / action | `Box(348,)` / `Box(17,)`, float32 |
+### One retention per episode
+Every episode carries a distinct seed and a **different retention** —
+`kv_cache_max_len` from 1 to 1000, each value used exactly once. Retention is
+how many steps of its own past the policy keeps while acting. It is a load-time
+argument, so all 1000 episodes come from the same weights and the same trained
+context length of 32; only how much history each rollout kept differs.
+`episode_metadata.jsonl` maps every episode file to the retention that recorded
+it — episodes are numbered in the order they finished, not by retention:
+```json
+{"episode": "ep_000000", "row": 4, "transitions": 113, "terminated": true, "truncated": false, "kv_cache_max_len": 16}
+```
+That makes the dataset usable two ways: as offline-RL trajectories, and as a
+measurement of what retention does to a policy's behaviour. Within it, episodes
+recorded between 16 and 128 steps of retention average 91.1 normalized against
+85.5 for those above 500 — but with one episode per retention, seed variance is
+larger than the effect, so treat the per-retention axis as evidence to build on
+rather than a settled result.
+### Files
+```
+humanoid/medium-v0/
+  data/main_data.hdf5        the Minari dataset
+  data/metadata.json
+  episode_metadata.jsonl     one line per episode: file, row, length, retention
+  collection_spec.json       what recorded it: bundle, context length, retentions
+  validation_summary.json    every number quoted above
+```
+### Load it
+```python
+import minari
+dataset = minari.load_dataset("ccnets/humanoid/medium-v0")
+episode = next(dataset.iterate_episodes())
+print(episode.observations.shape, episode.rewards.sum())
+```
+Place the `humanoid/medium-v0` directory under `ccnets/` in your Minari
+datasets path — `~/.minari/datasets/ccnets/humanoid/medium-v0` — or point
+`MINARI_DATASETS_PATH` at wherever you keep it.
+The files were recorded with `torch 2.12.0`, `gymnasium 1.2.2` and
+`mujoco 3.8.1`; a different simulator release is a different measurement even
+with identical weights and seeds. The recorder and packager are
+[`collection/`](https://github.com/ccnets-team/causal-gpt-rl/tree/main/collection),
+and the cycle this repository is built on is described in
+[Improving the next dataset](https://github.com/ccnets-team/causal-gpt-rl/blob/main/collection/docs/improving-the-next-dataset.md).
+## A note on the name
+`ccnets/humanoid/medium-v0` keeps the rung — `medium-v0` — and changes the
+namespace. The rung says where in the ladder the data was produced: after the
+medium trajectories, before expert. The namespace says who produced it, and
+keeps it clear of Farama's `mujoco/humanoid/medium-v0`, which is the dataset
+the recording policy was trained on. Minari resolves a dataset by id, so the two
+can sit in one `~/.minari` without either hiding the other — which matters
+here, because they are the input and the output of the same experiment.
+## Companion repos
+- Policy bundles for eight MuJoCo environments:
+  [ccnets/causal-gpt-rl](https://huggingface.co/ccnets/causal-gpt-rl)
+- Unity ML-Agents trajectories, with a calibrated quality ladder:
+  [ccnets/causal-gpt-rl-unity-datasets](https://huggingface.co/datasets/ccnets/causal-gpt-rl-unity-datasets)
+- Code:
+  [ccnets-team/causal-gpt-rl](https://github.com/ccnets-team/causal-gpt-rl)
+**한국어 카드 / Korean edition:**
+[README.ko.md](https://huggingface.co/datasets/ccnets/causal-gpt-rl-datasets/blob/main/README.ko.md)