AdrianLlopart commited on
Commit
e89633e
·
verified ·
1 Parent(s): 57f44c5

docs: HF model card for OpenRAL/rskill-molmoact2-multi-so101-nf4 v0.1.0

Browse files
Files changed (1) hide show
  1. README.md +7 -7
README.md CHANGED
@@ -23,7 +23,7 @@ base_model_relation: quantized
23
  inference: false
24
  ---
25
 
26
- # rskill-molmoact2-so101-nf4
27
 
28
  > **OpenRAL rSkill** — MolmoAct2 (Ai2's open action reasoning model: a
29
  > Molmo2-ER embodied-reasoning VLM backbone with a flow-matching
@@ -33,7 +33,7 @@ inference: false
33
  > Robots: SO-100 and SO-101 follower arms. **Apache-2.0 weights** — commercial
34
  > use permitted.
35
 
36
- This package wraps `hf://OpenRAL/rskill-molmoact2-so101-nf4` (an
37
  NF4-quantized mirror of `allenai/MolmoAct2-SO100_101`) with a `rskill.yaml`
38
  manifest that adds capability checking, license surfacing, latency budgets,
39
  and local registry integration. It does **not** copy model weights — they
@@ -128,13 +128,13 @@ it does **not** retrain or copy the full-precision weights.
128
 
129
  | Field | Value |
130
  | --- | --- |
131
- | `name` | `OpenRAL/rskill-molmoact2-so101-nf4` |
132
  | `version` | `0.1.0` |
133
  | `license` | `apache-2.0` |
134
  | `role` | `s1` |
135
  | `embodiment_tags` | `["so100_follower", "so101_follower"]` |
136
  | `runtime` / `quantization.dtype` | `pytorch` / `int4` (NF4) |
137
- | `weights_uri` | `hf://OpenRAL/rskill-molmoact2-so101-nf4` |
138
  | `chunk_size` / `n_action_steps` | 10 / 10 (full chunk replay) |
139
  | `latency_budget.per_chunk_ms` | 1000 ms |
140
  | `commercial_use_allowed` | `true` (Apache-2.0) |
@@ -154,7 +154,7 @@ print(pkg.manifest.name, pkg.manifest.version)
154
 
155
  ```bash
156
  # CLI:
157
- uv run openral rskill install OpenRAL/rskill-molmoact2-so101-nf4
158
  uv run openral rskill check # does this host meet the requirements?
159
  ```
160
 
@@ -201,7 +201,7 @@ host):
201
  ```bash
202
  HF_TOKEN=<write-token> uv run python tools/quantize_rskill.py \
203
  --source allenai/MolmoAct2-SO100_101 \
204
- --target OpenRAL/rskill-molmoact2-so101-nf4 \
205
  --loader transformers --trust-remote-code
206
  ```
207
 
@@ -215,7 +215,7 @@ numbers are available yet. Run the reproduction command in
215
 
216
  This rSkill package (`rskill.yaml`, `README.md`, `eval/so101.json`) is
217
  **Apache-2.0**. The wrapped weights at
218
- `hf://OpenRAL/rskill-molmoact2-so101-nf4` (NF4 mirror of
219
  `allenai/MolmoAct2-SO100_101`) are also released under **Apache-2.0** by Ai2 —
220
  commercial use is permitted; review the upstream LICENSE before deployment.
221
 
 
23
  inference: false
24
  ---
25
 
26
+ # rskill-molmoact2-multi-so101-nf4
27
 
28
  > **OpenRAL rSkill** — MolmoAct2 (Ai2's open action reasoning model: a
29
  > Molmo2-ER embodied-reasoning VLM backbone with a flow-matching
 
33
  > Robots: SO-100 and SO-101 follower arms. **Apache-2.0 weights** — commercial
34
  > use permitted.
35
 
36
+ This package wraps `hf://OpenRAL/rskill-molmoact2-multi-so101-nf4` (an
37
  NF4-quantized mirror of `allenai/MolmoAct2-SO100_101`) with a `rskill.yaml`
38
  manifest that adds capability checking, license surfacing, latency budgets,
39
  and local registry integration. It does **not** copy model weights — they
 
128
 
129
  | Field | Value |
130
  | --- | --- |
131
+ | `name` | `OpenRAL/rskill-molmoact2-multi-so101-nf4` |
132
  | `version` | `0.1.0` |
133
  | `license` | `apache-2.0` |
134
  | `role` | `s1` |
135
  | `embodiment_tags` | `["so100_follower", "so101_follower"]` |
136
  | `runtime` / `quantization.dtype` | `pytorch` / `int4` (NF4) |
137
+ | `weights_uri` | `hf://OpenRAL/rskill-molmoact2-multi-so101-nf4` |
138
  | `chunk_size` / `n_action_steps` | 10 / 10 (full chunk replay) |
139
  | `latency_budget.per_chunk_ms` | 1000 ms |
140
  | `commercial_use_allowed` | `true` (Apache-2.0) |
 
154
 
155
  ```bash
156
  # CLI:
157
+ uv run openral rskill install OpenRAL/rskill-molmoact2-multi-so101-nf4
158
  uv run openral rskill check # does this host meet the requirements?
159
  ```
160
 
 
201
  ```bash
202
  HF_TOKEN=<write-token> uv run python tools/quantize_rskill.py \
203
  --source allenai/MolmoAct2-SO100_101 \
204
+ --target OpenRAL/rskill-molmoact2-multi-so101-nf4 \
205
  --loader transformers --trust-remote-code
206
  ```
207
 
 
215
 
216
  This rSkill package (`rskill.yaml`, `README.md`, `eval/so101.json`) is
217
  **Apache-2.0**. The wrapped weights at
218
+ `hf://OpenRAL/rskill-molmoact2-multi-so101-nf4` (NF4 mirror of
219
  `allenai/MolmoAct2-SO100_101`) are also released under **Apache-2.0** by Ai2 —
220
  commercial use is permitted; review the upstream LICENSE before deployment.
221