This is a project done in google colab. I’m given a .hdf5 file with robot trajectory demonstrations. The assignment is to film new trajectories using architectures we’ve learned about in class. I’ve done that, it’s the copying the new trajectory to a new .hdf5 file that seems to be an issue

def segment_data(f, filename, num_demos, testing):
    keys = list(f["data"].keys())  # get all demonstration ids

    for i in range(1, num_demos):  # iterate over all demonstrations
        subfile = filename + ".hdf5"
        if os.path.exists(subfile):
            os.remove(subfile)
        f_new = h5py.File(subfile, "w")
        grp = f_new.create_group("data")
        grp.attrs.create("env_args", f["data"].attrs["env_args"])  # copy over the environment args
        grp.attrs.create("total", f["data"].attrs["total"])  # copy over the "total" attribute

        # copy over each of the next demonstrations to the new group
        f.copy("data/{}".format(keys[i]), grp, name=keys[i])

        # edit 'obs' data if it exists
        if 'obs' in grp[keys[i]]:
            # Modify 'robot_eef_pos' data here
            grp[keys[i]]['obs']['robot0_eef_pos'][:] = [[1.1, 2.1, 3.1] for _ in range(len(grp[keys[i]]['obs']['robot0_eef_pos']))]
            grp[keys[i]]['obs']['robot0_gripper_qpos'][:] = [[0.0, 0.0] for _ in range(len(grp[keys[i]]['obs']['robot0_gripper_qpos']))]

        

        f_new.close()

f is a pointer pointing to the file that contains the robot trajectories, filename is the file I want to save it to, num_demos is the number of demonstrations in f (yes I know this code will overwrite every bit of data except the last demonstration. I plan to fix that later) testing is well, the new demonstrations I want to test. But currently I’m trying to make it so that ‘obs’ basically stays constant for every end effector position and gripper position. When I print out. When I have a pointer pointing to the new file and asking it what the values are, it correctly says the constants I put in, but when I use the function to play the video, the trajectory is identical used as a demonstration