fix rendering bug and update eval scripts

liuzuxin · Aug 25, 2023 · f34d324 · f34d324
1 parent ad48ac8
commit f34d324
Show file tree

Hide file tree

Showing 8 changed files with 21 additions and 9 deletions.
diff --git a/examples/mlp/eval_cpo_agent.py b/examples/mlp/eval_cpo_agent.py
@@ -48,11 +48,13 @@ def eval(args: EvalConfig):
  )
 
  if args.parallel_eval:
+ assert args.render is False, "please use single env when rendering"
  test_envs = ShmemVectorEnv(
  [lambda: gym.make(task) for _ in range(args.eval_episodes)]
  )
  else:
- test_envs = gym.make(task)
+ render_mode = "human" if args.render else None
+ test_envs = gym.make(task, render_mode=render_mode)
 
  rews, lens, cost = agent.evaluate(
  test_envs=test_envs,

diff --git a/examples/mlp/eval_cvpo_agent.py b/examples/mlp/eval_cvpo_agent.py
@@ -47,11 +47,13 @@ def eval(args: EvalConfig):
  )
 
  if args.parallel_eval:
+ assert args.render is False, "please use single env when rendering"
  test_envs = ShmemVectorEnv(
  [lambda: gym.make(task) for _ in range(args.eval_episodes)]
  )
  else:
- test_envs = gym.make(task)
+ render_mode = "human" if args.render else None
+ test_envs = gym.make(task, render_mode=render_mode)
 
  rews, lens, cost = agent.evaluate(
  test_envs=test_envs,

diff --git a/examples/mlp/eval_ddpgl_agent.py b/examples/mlp/eval_ddpgl_agent.py
@@ -47,11 +47,13 @@ def eval(args: EvalConfig):
  )
 
  if args.parallel_eval:
+ assert args.render is False, "please use single env when rendering"
  test_envs = ShmemVectorEnv(
  [lambda: gym.make(task) for _ in range(args.eval_episodes)]
  )
  else:
- test_envs = gym.make(task)
+ render_mode = "human" if args.render else None
+ test_envs = gym.make(task, render_mode=render_mode)
 
  rews, lens, cost = agent.evaluate(
  test_envs=test_envs,

diff --git a/examples/mlp/eval_focops_agent.py b/examples/mlp/eval_focops_agent.py
@@ -48,11 +48,13 @@ def eval(args: EvalConfig):
  )
 
  if args.parallel_eval:
+ assert args.render is False, "please use single env when rendering"
  test_envs = ShmemVectorEnv(
  [lambda: gym.make(task) for _ in range(args.eval_episodes)]
  )
  else:
- test_envs = gym.make(task)
+ render_mode = "human" if args.render else None
+ test_envs = gym.make(task, render_mode=render_mode)
 
  rews, lens, cost = agent.evaluate(
  test_envs=test_envs,

diff --git a/examples/mlp/eval_ppol_agent.py b/examples/mlp/eval_ppol_agent.py
@@ -48,11 +48,13 @@ def eval(args: EvalConfig):
  )
 
  if args.parallel_eval:
+ assert args.render is False, "please use single env when rendering"
  test_envs = ShmemVectorEnv(
  [lambda: gym.make(task) for _ in range(args.eval_episodes)]
  )
  else:
- test_envs = gym.make(task)
+ render_mode = "human" if args.render else None
+ test_envs = gym.make(task, render_mode=render_mode)
 
  rews, lens, cost = agent.evaluate(
  test_envs=test_envs,

diff --git a/examples/mlp/eval_sacl_agent.py b/examples/mlp/eval_sacl_agent.py
@@ -48,11 +48,13 @@ def eval(args: EvalConfig):
  )
 
  if args.parallel_eval:
+ assert args.render is False, "please use single env when rendering"
  test_envs = ShmemVectorEnv(
  [lambda: gym.make(task) for _ in range(args.eval_episodes)]
  )
  else:
- test_envs = gym.make(task)
+ render_mode = "human" if args.render else None
+ test_envs = gym.make(task, render_mode=render_mode)
 
  rews, lens, cost = agent.evaluate(
  test_envs=test_envs,

diff --git a/examples/mlp/eval_trpo_agent.py b/examples/mlp/eval_trpo_agent.py
@@ -48,11 +48,13 @@ def eval(args: EvalConfig):
  )
 
  if args.parallel_eval:
+ assert args.render is False, "please use single env when rendering"
  test_envs = ShmemVectorEnv(
  [lambda: gym.make(task) for _ in range(args.eval_episodes)]
  )
  else:
- test_envs = gym.make(task)
+ render_mode = "human" if args.render else None
+ test_envs = gym.make(task, render_mode=render_mode)
 
  rews, lens, cost = agent.evaluate(
  test_envs=test_envs,

diff --git a/fsrl/data/fast_collector.py b/fsrl/data/fast_collector.py
@@ -328,8 +328,6 @@ def collect(
 
  if render:
  self.env.render()
- if "render_fps" not in self.env.metadata:
- time.sleep(0.1)
 
  # add data into the buffer
  ptr, ep_rew, ep_len, ep_idx = self.buffer.add(