Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →mkurman/metaworld
Robotics multi-task and meta-reinforcement learning benchmark (Meta-World). Standardized continuous-control benchmark built on Gymnasium with MT1, MT10, MT50 multi-task suites and ML1, ML10, ML45 meta-learning suites. Supports synchronous and asynchronous vector execution.
Passed 20 of 22 test cases with the skill, and 10 of 22 without it.
and 15 more in the full report
12 fixed8 passed either way0 still failing2 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 12 test cases the skill fixed. Same model, same prompt, with and without the skill.
I need to set up a broad 50-task multi-task manipulation benchmark environment using Gymnasium's vectorized environment creation function. Which benchmark ID should I use?
| Checked | Without skill | With skill |
|---|---|---|
| Uses benchmark environment ID 'Meta-World/MT50'. | Missed | Passed |
| Calls gym.make_vec to create the vectorized multi-task environment. | Missed | Passed |
The instructions your model reads when the skill loads.
-----|---------|------|
| MT1 | Multi-task learning on one selected task family | 1 task |
| MT10 | Multi-task learning across 10 manipulation tasks | 10 tasks |
| MT50 | Broad multi-task benchmark | 50 tasks |
| ML1-train/test | Meta-learning goal variation within one task family | 1 family |
| ML10-train/test | Meta-learning across train/test task split | 10 train + 5 test |
| ML45-train/test | Large-scale meta-learning split | 45 train + 5 test |
MT1:
import gymnasium as gym
import metaworld
env = gym.make("Meta-World/MT1", env_name="reach-v3", seed=42)
obs, info = env.reset()
action = env.action_space.sample()
obs, reward, terminated, truncated, info = env.step(action)
MT10 synchronous vectorized:
import gymnasium as gym
import metaworld
envs = gym.make_vec("Meta-World/MT10", vector_strategy="sync", seed=42)
obs, info = envs.reset()
actions = envs.action_space.sample()
obs, rewards, terminations, truncations, infos = envs.step(actions)
MT10 asynchronous vectorized:
envs = gym.make_vec("Meta-World/MT10", vector_strategy="async", seed=42)
MT50:
envs = gym.make_vec("Meta-World/MT50", vector_strategy="sync", seed=42)
test cases passed, vs 10 / 22 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.