Fix render arguments when using single tasks

The current implementation has an issue in that it only allows changing render_mode during rendering

. However, in many cases, we need to change rendered image size as well as camera view. I replace render_mode with **render_kwargs such that users can set width height and camera_name when initializing the environment.

Modifications

• metaworld/envs/mujoco/sawyer_xyz/v2/sawyer_*.py are modified. e.g.,

  Metaworld/metaworld/envs/mujoco/sawyer_xyz/v2/sawyer_assembly_peg_v2.py

  Lines 20 to 22 in cca35cf

  	render_mode: RenderMode | None = None,
  	camera_name: str | None = None,
  	camera_id: int | None = None,

  and

  Metaworld/metaworld/envs/mujoco/sawyer_xyz/v2/sawyer_assembly_peg_v2.py

  Lines 34 to 36 in cca35cf

  	render_mode=render_mode,
  	camera_name=camera_name,
  	camera_id=camera_id,

  are replaced with **render_kwargs.
• metaworld/envs/mujoco/sawyer_xyz/sawyer_xyz_env.py is modified the same as above.
• metaworld/envs/mujoco/env_dict/py is modified. render_mode is replaced with **render_kwargs in

  Metaworld/metaworld/envs/mujoco/env_dict.py

  Line 182 in cca35cf

  def initialize(env, seed=None, render_mode=None):

  , and

  Metaworld/metaworld/envs/mujoco/env_dict.py

  Line 191 in cca35cf

  env.render_mode = render_mode

  is removed.

Additional dependencies

None

TODO:

Currently, for camera corner, corner2 and corner3, the rendered images incorrectly rotate 90 degrees. It might be because of using OpenGL vs. opencv as the rendering backbone. image[::-1] can correct the images when using these camera views.