Publish paired asset 0.1.6 and configuration 0.2.6 manifests and guide
Browse files- README.md +23 -13
- assets/0.1.6/sha256-manifest.json +0 -0
- saved-instances/0.2.6/sha256-manifest.json +0 -0
README.md
CHANGED
|
@@ -13,7 +13,7 @@ tags:
|
|
| 13 |
|
| 14 |
# RobotAgentBench simulation assets
|
| 15 |
|
| 16 |
-
Version 0.1.
|
| 17 |
Context Reasoning question banks and standard answers for RobotAgentBench.
|
| 18 |
Evaluation code: https://github.com/AetherLabsAI/RobotAgentBench/tree/main.
|
| 19 |
|
|
@@ -69,14 +69,23 @@ reused unchanged. Update the evaluator, reuse verified cached downloads,
|
|
| 69 |
and install into fresh asset and runtime directories. Use configuration
|
| 70 |
bank **0.2.5** with this release.
|
| 71 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 72 |
## Install
|
| 73 |
|
| 74 |
Install the evaluator from its code repository, then use its versioned manifest:
|
| 75 |
|
| 76 |
```sh
|
| 77 |
-
robotagentbench assets fetch resources/assets-0.1.
|
| 78 |
-
robotagentbench assets install resources/assets-0.1.
|
| 79 |
-
robotagentbench assets configure resources/assets-0.1.
|
| 80 |
```
|
| 81 |
|
| 82 |
For offline installation, copy all release archive parts into `downloads`.
|
|
@@ -92,7 +101,7 @@ The separate configuration release contains **1,470 saved initial configurations
|
|
| 92 |
reproducible evaluation, not the exact instances used for the paper's reported
|
| 93 |
results. Each record retains its saved seed and initial state.
|
| 94 |
|
| 95 |
-
Configuration release **0.2.
|
| 96 |
of the four OpenEnd scenarios. Household memory and the high-level request
|
| 97 |
are accompanied by **Today's situation**, covering weather, trip timing,
|
| 98 |
confirmed plans, and personal preferences.
|
|
@@ -103,19 +112,20 @@ Variants that differ only in visible state use the same public text. The six
|
|
| 103 |
context/scene variants per scenario each have five spatial initial states;
|
| 104 |
they are not thirty distinct stories.
|
| 105 |
|
| 106 |
-
Version 0.2.
|
| 107 |
-
replacements that passed native initial admission
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
|
|
|
| 111 |
Task-owned attachments, original spawn receipts, and lossless room patches
|
| 112 |
reconstruct the saved configurations. Room meshes and textures remain in
|
| 113 |
the shared asset download.
|
| 114 |
-
The configuration download is approximately **
|
| 115 |
|
| 116 |
Earlier releases remain available under `saved-instances/0.1.0/` and
|
| 117 |
`saved-instances/0.2.0/`, `saved-instances/0.2.1/`, `saved-instances/0.2.2/`, and
|
| 118 |
-
`saved-instances/0.2.3/`, and `saved-instances/0.2.
|
| 119 |
|
| 120 |
Install the shared assets once. Configurations store task parameters, poses,
|
| 121 |
robot initialization, required cloth-state arrays, and compact lossless scene
|
|
@@ -123,7 +133,7 @@ patches; they do not duplicate complete room geometry and textures for every epi
|
|
| 123 |
|
| 124 |
```sh
|
| 125 |
python -m robotagentbench.simulation.saved_instance_release.download \
|
| 126 |
-
--manifest resources/saved-instances-0.2.
|
| 127 |
--cache downloads/configurations --destination /absolute/path/to/SavedConfigurations
|
| 128 |
```
|
| 129 |
|
|
|
|
| 13 |
|
| 14 |
# RobotAgentBench simulation assets
|
| 15 |
|
| 16 |
+
Version 0.1.6 supplies simulation assets, initial-state resources and the full
|
| 17 |
Context Reasoning question banks and standard answers for RobotAgentBench.
|
| 18 |
Evaluation code: https://github.com/AetherLabsAI/RobotAgentBench/tree/main.
|
| 19 |
|
|
|
|
| 69 |
and install into fresh asset and runtime directories. Use configuration
|
| 70 |
bank **0.2.5** with this release.
|
| 71 |
|
| 72 |
+
## Shared asset release 0.1.6
|
| 73 |
+
|
| 74 |
+
Use this release with saved configuration bank **0.2.6**. It adds one compact,
|
| 75 |
+
lossless shared scene-reconstruction dependency for the new Sofa recovery
|
| 76 |
+
configuration. The supplement is approximately **515 KB**; all 31 earlier
|
| 77 |
+
archives, all 50 room layouts, and third-party notices remain unchanged.
|
| 78 |
+
The complete asset download is approximately **7.04 GB** across **32 archives**.
|
| 79 |
+
Reuse verified downloads and install into fresh asset/runtime directories.
|
| 80 |
+
|
| 81 |
## Install
|
| 82 |
|
| 83 |
Install the evaluator from its code repository, then use its versioned manifest:
|
| 84 |
|
| 85 |
```sh
|
| 86 |
+
robotagentbench assets fetch resources/assets-0.1.6.json --cache downloads
|
| 87 |
+
robotagentbench assets install resources/assets-0.1.6.json --cache downloads --destination Assets
|
| 88 |
+
robotagentbench assets configure resources/assets-0.1.6.json --destination Assets --runtime Runtime --profile room
|
| 89 |
```
|
| 90 |
|
| 91 |
For offline installation, copy all release archive parts into `downloads`.
|
|
|
|
| 101 |
reproducible evaluation, not the exact instances used for the paper's reported
|
| 102 |
results. Each record retains its saved seed and initial state.
|
| 103 |
|
| 104 |
+
Configuration release **0.2.6** contains 30 saved configurations for each
|
| 105 |
of the four OpenEnd scenarios. Household memory and the high-level request
|
| 106 |
are accompanied by **Today's situation**, covering weather, trip timing,
|
| 107 |
confirmed plans, and personal preferences.
|
|
|
|
| 112 |
context/scene variants per scenario each have five spatial initial states;
|
| 113 |
they are not thirty distinct stories.
|
| 114 |
|
| 115 |
+
Version 0.2.6 updates three saved initial states using original-rule seeded
|
| 116 |
+
replacements that passed native initial admission: lighting episode 4,
|
| 117 |
+
combined coffee disturbance episode 5, and sofa recovery episode 2. The
|
| 118 |
+
other 1,467 states, public contexts, task definitions, scoring rules, and
|
| 119 |
+
physical parameters are unchanged from 0.2.5; its two earlier replacements
|
| 120 |
+
are retained. Earlier versions remain available.
|
| 121 |
Task-owned attachments, original spawn receipts, and lossless room patches
|
| 122 |
reconstruct the saved configurations. Room meshes and textures remain in
|
| 123 |
the shared asset download.
|
| 124 |
+
The configuration download is approximately **250 MB**, split into seven archives.
|
| 125 |
|
| 126 |
Earlier releases remain available under `saved-instances/0.1.0/` and
|
| 127 |
`saved-instances/0.2.0/`, `saved-instances/0.2.1/`, `saved-instances/0.2.2/`, and
|
| 128 |
+
`saved-instances/0.2.3/`, `saved-instances/0.2.4/`, and `saved-instances/0.2.5/`. Install 0.2.6 into a new directory when upgrading.
|
| 129 |
|
| 130 |
Install the shared assets once. Configurations store task parameters, poses,
|
| 131 |
robot initialization, required cloth-state arrays, and compact lossless scene
|
|
|
|
| 133 |
|
| 134 |
```sh
|
| 135 |
python -m robotagentbench.simulation.saved_instance_release.download \
|
| 136 |
+
--manifest resources/saved-instances-0.2.6.json \
|
| 137 |
--cache downloads/configurations --destination /absolute/path/to/SavedConfigurations
|
| 138 |
```
|
| 139 |
|
assets/0.1.6/sha256-manifest.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
saved-instances/0.2.6/sha256-manifest.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|