lerobot/Cosmos3-Edge-Policy-LIBERO
Overview
This release provides EMA weights for a Cosmos3-Edge policy fine-tuned on four LIBERO suites (Spatial, Object, Goal, and LIBERO-10). The model predicts 16 actions and future video frames from third-person and wrist camera images. Weights correspond to optimizer update 4000 of the training run. Evaluation was conducted with MuJoCo 2.3.7 on NVIDIA A800 GPUs; results are not comparable with MuJoCo 3.x due to contact physics changes. The repository includes training commands, evaluation scripts, and license notices for redistribution.
Generated from attributed reports · Updated 1 hours ago
Event evidence and corrections
1 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Dataset · license: OpenMDW 1.1
“-Edge-Policy-LIBERO loads the configuration, the weights and the processors. config.json sets dtype: float32 on purpose: fine-tuning from this checkpoint keeps float32 master weights. For inference and evaluation pass --policy.dtype=bfloat16 . License and provenance The NVIDIA model materials use OpenMDW 1.1 . Retain the license and applicable copyright and origin notices when redistributing them. NOTICE records the”
Exact source · revision 2huggingface
Dataset · tasks: LIBERO-Spatial · Differing source assertions
“cessors Results on LIBERO Model Libero-Spatial Libero-Object Libero-Goal Libero-10 Average Cosmos3-Nano-Policy-LIBERO 98.0% (490/500) 99.4% (497/500) 97.4% (487/500) 96.8% (484/500) 97.90% Cosmos3-Edge-Policy-LIBERO 98.6% (493/500) 99.4% (497/500) 97.2% (486/500) 96.6% (483/500) 97.95% Closed-loop success rates: 50 episodes for each of the 10 tasks of a suite, 500 episodes per suite, with the evaluation setup below.”
Exact source · revision 2huggingface
Dataset · tasks: LIBERO-Object · Differing source assertions
“cessors Results on LIBERO Model Libero-Spatial Libero-Object Libero-Goal Libero-10 Average Cosmos3-Nano-Policy-LIBERO 98.0% (490/500) 99.4% (497/500) 97.4% (487/500) 96.8% (484/500) 97.90% Cosmos3-Edge-Policy-LIBERO 98.6% (493/500) 99.4% (497/500) 97.2% (486/500) 96.6% (483/500) 97.95% Closed-loop success rates: 50 episodes for each of the 10 tasks of a suite, 500 episodes per suite, with the evaluation setup below.”
Exact source · revision 2huggingface
Dataset · tasks: LIBERO-Goal · Differing source assertions
“cessors Results on LIBERO Model Libero-Spatial Libero-Object Libero-Goal Libero-10 Average Cosmos3-Nano-Policy-LIBERO 98.0% (490/500) 99.4% (497/500) 97.4% (487/500) 96.8% (484/500) 97.90% Cosmos3-Edge-Policy-LIBERO 98.6% (493/500) 99.4% (497/500) 97.2% (486/500) 96.6% (483/500) 97.95% Closed-loop success rates: 50 episodes for each of the 10 tasks of a suite, 500 episodes per suite, with the evaluation setup below.”
Exact source · revision 2huggingface
Dataset · tasks: LIBERO-10 · Differing source assertions
“cessors Results on LIBERO Model Libero-Spatial Libero-Object Libero-Goal Libero-10 Average Cosmos3-Nano-Policy-LIBERO 98.0% (490/500) 99.4% (497/500) 97.4% (487/500) 96.8% (484/500) 97.90% Cosmos3-Edge-Policy-LIBERO 98.6% (493/500) 99.4% (497/500) 97.2% (486/500) 96.6% (483/500) 97.95% Closed-loop success rates: 50 episodes for each of the 10 tasks of a suite, 500 episodes per suite, with the evaluation setup below.”
Exact source · revision 2huggingface
Dataset · embodiments: LeRobot policy
“Cosmos3-Edge fine-tuned on the four LIBERO suites (Spatial, Object, Goal and LIBERO-10) as a LeRobot policy. From the task instruction and the third person and wrist camera images, the policy predicts a chunk of 16 actions together with the future video of the scene. This repository holds the EMA weights at optimizer update 4000. Item Value Base checkpoint lerobot/Cosmos3-Edge , the LeRobot format of nvidia/Cosmos3-E”
Exact source · revision 2huggingface
Report timeline
Follow attributed reports and material updates.
- LeRobot — Model repository updatesSignallerobot/Cosmos3-Edge-Policy-LIBERO
This release provides EMA weights for a Cosmos3-Edge policy fine-tuned on four LIBERO suites (Spatial, Object, Goal, and LIBERO-10). The model predicts 16 actions and future video frames from third-person and wrist camera images. Weights correspond to optimizer update 4000 of the training run. Evaluation was conducted with MuJoCo 2.3.7 on NVIDIA A800 GPUs; results are not comparable with MuJoCo 3.x due to contact physics changes. The repository includes training commands, evaluation scripts, and license notices for redistribution.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).