Open tanxr00 opened 6 months ago
Hey, maybe you could look at replay_episodes.py
? It has the following code:
def main(args):
# dataset_path is the file path to a .h5py file that contains
# action (states of the various components) which we then sequentially perform
dataset_path = args['dataset_path']
with h5py.File(dataset_path, 'r') as root:
actions = root['/action'][()]
# load the environment fo the task:
env = make_sim_env('sim_transfer_cube')
BOX_POSE[0] = sample_box_pose() # used in sim reset
ts = env.reset()
episode_replay = [ts]
# perform the action-
for action in actions:
print("For action:")
print(action)
ts = env.step(action)
episode_replay.append(ts)
A single action may look like-
[-1.2901543e-03 -3.6779362e-01 1.1062293e+00 -2.0220939e-03 -7.3979789e-01 1.4955290e-03 0.0000000e+00 3.1307980e-05 -2.9615888e-01 3.0987167e-01 -3.2766675e-05 1.0336729e+00 5.0244853e-05 1.0000000e+00]
Hi, thank you for great work.
I get a issue. How to real-time inference?
or infer a trained scene, can infer a new scene?