diff --git a/envs/batched_env.py b/envs/batched_env.py
index 796e29641a7299145993375306f8cd1331bd71b1..2543d6bc30890b01822d050779b9a87cef99d81c 100644
--- a/envs/batched_env.py
+++ b/envs/batched_env.py
@@ -42,13 +42,6 @@ class BatchedEnv:
observation = [env.reset() for env in self.envs]
return observation
- def single_env_reset(self, index):
- """
- Resets the env at the index location
- """
- observation = self.envs[index].reset()
- return observation
-
if __name__ == '__main__':
diff --git a/local_evaluation.py b/local_evaluation.py
deleted file mode 100644
index 5da924a72f2da615824d021aa7055bdd008f9fa5..0000000000000000000000000000000000000000
--- a/local_evaluation.py
+++ /dev/null
@@ -1,38 +0,0 @@
-## This file is intended to emulate the evaluation on AIcrowd
-
-# IMPORTANT - Differences to expect
-# * All the environment's functions are not available
-# * The run might be slower than your local run
-# * Resources might vary from your local machine
-
-from submission_agent import SubmissionConfig, LocalEvaluationConfig
-
-from rollout import run_batched_rollout
-from nethack_baselines.utils.batched_env import BatchedEnv
-
-
-# Ideally you shouldn't need to change anything below
-def add_evaluation_wrappers_fn(env_make_fn):
- max_episodes = LocalEvaluationConfig.LOCAL_EVALUATION_NUM_EPISODES
- # TOOD: use LOCAL_EVALUATION_NUM_EPISODES for limiting episodes
- return env_make_fn
-
-def evaluate():
- submission_env_make_fn = SubmissionConfig.submission_env_make_fn
- num_envs = SubmissionConfig.NUM_PARALLEL_ENVIRONMENTS
- Agent = SubmissionConfig.Submision_Agent
-
- evaluation_env_fn = add_evaluation_wrappers_fn(submission_env_make_fn)
- batched_env = BatchedEnv(env_make_fn=evaluation_env_fn,
- num_envs=num_envs)
-
- num_envs = batched_env.num_envs
- num_actions = batched_env.num_actions
-
- agent = Agent(num_envs, num_actions)
-
- run_batched_rollout(batched_env, agent)
-
-
-if __name__ == '__main__':
- evaluate()
diff --git a/rollout.py b/rollout.py
index 586c4b22d29d70563d0bc8c34a00ec96ff58dd60..aac7720dd9aa636e5d075cc79f3f983bde895d6b 100644
--- a/rollout.py
+++ b/rollout.py
@@ -10,11 +10,12 @@ from tqdm import tqdm
import numpy as np
from envs.batched_env import BatchedEnv
+from envs.wrappers import create_env
from submission_config import SubmissionConfig
-NUM_ASSESSMENTS = 512
-def run_batched_rollout(batched_env, agent):
+
+def run_batched_rollout(num_episodes, batched_env, agent):
"""
This function will generate a series of rollouts in a batched manner.
"""
@@ -28,16 +29,16 @@ def run_batched_rollout(batched_env, agent):
infos = [{} for _ in range(num_envs)]
# We mark at the start of each episode if we are 'counting it'
- active_envs = [i < NUM_ASSESSMENTS for i in range(num_envs)]
- num_remaining = NUM_ASSESSMENTS - sum(active_envs)
+ active_envs = [i < num_episodes for i in range(num_envs)]
+ num_remaining = num_episodes - sum(active_envs)
episode_count = 0
- pbar = tqdm(total=NUM_ASSESSMENTS)
+ pbar = tqdm(total=num_episodes)
all_returns = []
returns = [0.0 for _ in range(num_envs)]
# The evaluator will automatically stop after the episodes based on the development/test phase
- while episode_count < NUM_ASSESSMENTS:
+ while episode_count < num_episodes:
actions = agent.batched_step(observations, rewards, dones, infos)
observations, rewards, dones, infos = batched_env.batch_step(actions)
@@ -57,20 +58,19 @@ def run_batched_rollout(batched_env, agent):
pbar.update(1)
returns[done_idx] = 0.0
+ pbar.close()
return all_returns
if __name__ == "__main__":
- submission_env_make_fn = SubmissionConfig.submission_env_make_fn
- NUM_PARALLEL_ENVIRONMENTS = SubmissionConfig.NUM_PARALLEL_ENVIRONMENTS
- Agent = SubmissionConfig.Submision_Agent
+ # AIcrowd will cut the assessment early duing the dev phase
+ NUM_ASSESSMENTS = 4096
- batched_env = BatchedEnv(
- env_make_fn=submission_env_make_fn, num_envs=NUM_PARALLEL_ENVIRONMENTS
- )
+ env_make_fn = SubmissionConfig.MAKE_ENV_FN
+ num_envs = SubmissionConfig.NUM_ENVIRONMENTS
+ Agent = SubmissionConfig.AGENT
- num_envs = batched_env.num_envs
- num_actions = batched_env.num_actions
- agent = Agent(num_envs, num_actions)
+ batched_env = BatchedEnv(env_make_fn=env_make_fn, num_envs=num_envs)
+ agent = Agent(num_envs, batched_env.num_actions)
- run_batched_rollout(batched_env, agent)
+ run_batched_rollout(NUM_ASSESSMENTS, batched_env, agent)
diff --git a/submission_config.py b/submission_config.py
index 5c28e9999cd55fb73717652dc03e10c4d65b3be8..9968e9996021f3040b04f47d7ec4b88dcf08b88d 100644
--- a/submission_config.py
+++ b/submission_config.py
@@ -15,26 +15,26 @@ from envs.wrappers import addtimelimitwrapper_fn
class SubmissionConfig:
## Add your own agent class
- Submision_Agent = TorchBeastAgent
- # Submision_Agent = RLlibAgent
- # Submision_Agent = RandomAgent
+ AGENT = TorchBeastAgent
+ # AGENT = RLlibAgent
+ # AGENT = RandomAgent
- ## Change the NUM_PARALLEL_ENVIRONMENTS as you need
+ ## Change the NUM_ENVIRONMENTS as you need
## for example reduce it if your GPU doesn't fit
## Increasing above 32 is not advisable for the Nethack Challenge 2021
- NUM_PARALLEL_ENVIRONMENTS = 32
+ NUM_ENVIRONMENTS = 32
## Add a function that creates your nethack env
## Mainly this is to add wrappers
## Add your wrappers to envs/wrappers.py and change the name here
## IMPORTANT: Don't "call" the function, only provide the name
- submission_env_make_fn = addtimelimitwrapper_fn
+ MAKE_ENV_FN = addtimelimitwrapper_fn
-class LocalEvaluationConfig:
+class TestEvaluationConfig:
# Change this to locally check a different number of rollouts
# The AIcrowd submission evaluator will not use this
# It is only for your local evaluation
- LOCAL_EVALUATION_NUM_EPISODES = 50
+ NUM_EPISODES = 64
diff --git a/test_submission.py b/test_submission.py
new file mode 100644
index 0000000000000000000000000000000000000000..7ab3495f27719d55fff6097f60df95ab429b1e87
--- /dev/null
+++ b/test_submission.py
@@ -0,0 +1,33 @@
+## This file is intended to emulate the evaluation on AIcrowd
+
+# IMPORTANT - Differences to expect
+# * All the environment's functions are not available
+# * The run might be slower than your local run
+# * Resources might vary from your local machine
+
+import numpy as np
+
+from agents.batched_agent import BatchedAgent
+from submission_config import SubmissionConfig, TestEvaluationConfig
+
+from rollout import run_batched_rollout
+from envs.batched_env import BatchedEnv
+
+
+def evaluate():
+ env_make_fn = SubmissionConfig.MAKE_ENV_FN
+ num_envs = SubmissionConfig.NUM_ENVIRONMENTS
+ Agent = SubmissionConfig.AGENT
+
+ num_episodes = TestEvaluationConfig.NUM_EPISODES
+
+ batched_env = BatchedEnv(env_make_fn=env_make_fn, num_envs=num_envs)
+
+ agent = Agent(num_envs, batched_env.num_actions)
+
+ scores = run_batched_rollout(num_episodes, batched_env, agent)
+ print(f"Median Score: {np.median(scores)}, Mean Score: {np.mean(scores)}")
+
+
+if __name__ == "__main__":
+ evaluate()