lets create trading

agents/trading_agent.py

@@ -9,3 +9,9 @@ class TradingAgent(BaseAgent):
     """
 
     name = "TradingAgent"
+
+
+class TradingAgentLogic():
+    """Logic for trading agent"""
+    def __init__(self,) -> None:
+        pass

components/simple_build.py

@@ -44,7 +44,7 @@ class SimpleCraft(BaseComponent):
         payment=10,
         payment_max_skill_multiplier=1,
         skill_dist="none",
-        build_labor=10.0,
+        build_labor=1.0,
         **base_component_kwargs
     ):
         super().__init__(*base_component_args, **base_component_kwargs)

main.py

@@ -57,9 +57,9 @@ env_config = {
     'allow_observation_scaling': True,
     'dense_log_frequency': 100, 
     'world_dense_log_frequency':1,
-   'energy_cost':0,
+    'energy_cost':0.21,
     'energy_warmup_method': "auto",
-    'energy_warmup_constant': 0,
+    'energy_warmup_constant': 4000,
     
     # In multi-action-mode, the policy selects an action for each action subspace (defined in component code).
     # Otherwise, the policy selects only 1 action.
@@ -105,13 +105,13 @@ eval_env_config = {
     # kwargs that are used by every Scenario class (i.e. defined in BaseEnvironment)
     'agent_composition': {"BasicMobileAgent": 20},          # Number of non-planner agents (must be > 1)
     'world_size': [1, 1], # [Height, Width] of the env world
-    'episode_length': 100, # Number of timesteps per episode
+    'episode_length': 256, # Number of timesteps per episode
     'allow_observation_scaling': True,
     'dense_log_frequency': 10, 
     'world_dense_log_frequency':1, 
-    'energy_cost':0,
+    'energy_cost':0.21,
     'energy_warmup_method': "auto",
-    'energy_warmup_constant': 0,
+    'energy_warmup_constant': 4000,
     
     # In multi-action-mode, the policy selects an action for each action subspace (defined in component code).
     # Otherwise, the policy selects only 1 action.
@@ -228,7 +228,7 @@ obs=monenv.reset()
 
 runname="run_{}".format(int(np.random.rand()*100))
 
-model = MaskablePPO("MlpPolicy",n_steps=int(env_config['episode_length']*2),ent_coef=0.1, vf_coef=0.8 ,gamma=0.95, learning_rate=5e-3,env=monenv, verbose=1,device="cuda",tensorboard_log="./log")
+model = MaskablePPO("MlpPolicy",n_steps=int(env_config['episode_length']*2),ent_coef=0.1, vf_coef=0.8 ,gamma=0.95, learning_rate=5e-3,env=monenv, seed=225,verbose=1,device="cuda",tensorboard_log="./log")
 n_agents=econ.n_agents
 total_required_for_episode=n_agents*env_config['episode_length']
 print("this is run {}".format(runname))
@@ -242,7 +242,7 @@ while True:
     # Create Eval ENV
     vec_mon_eval=VecMonitor(venv=eval_sb3_converter)
     #Train
-    model=model.learn(total_timesteps=total_required_for_episode*20,progress_bar=True,reset_num_timesteps=False,tb_log_name=runname,callback=TensorboardCallback(econ=econ))
+    model=model.learn(total_timesteps=total_required_for_episode*10,progress_bar=True,reset_num_timesteps=False,tb_log_name=runname,callback=TensorboardCallback(econ=econ))
     #normenv.save("temp-normalizer.ai")
 
     
@@ -253,7 +253,9 @@ while True:
     obs=vec_mon_eval.reset()
     done=False
     for i in tqdm(range(eval_env_config['episode_length'])):
-        action=model.predict(obs)
+        masks=eval_sb3_converter.action_masks()
+        action=model.predict(obs,action_masks=masks)
+
         obs,rew,done_e,info=vec_mon_eval.step(action[0])
         done=done_e[0]
 

wrapper/sb3_econ_converter.py

@@ -82,16 +82,18 @@ class SB3EconConverter(VecEnv, gym.Env):
     def seed(self, seed: Optional[int] = None) -> List[Union[None, int]]:
         if seed is None:
             seed = np.random.randint(0, 2**32 - 1)
-        seeds = []
-        for idx, env in enumerate(self.envs):
-            seeds.append(env.seed(seed + idx))
+        self.econ.seed(seed)
+        seeds=[seed]
         return seeds
 
     def action_masks(self):
         """Returns action masks for agents and current obs"""
         masks=[]
         for obs in self.curr_obs:
-            masks.append(self.curr_obs[obs]["action_mask"])
+            mask=[]
+            for num in self.curr_obs[obs]["action_mask"]:
+                mask.append(num==1.0)
+            masks.append(mask)
         return masks
 
     def close(self) -> None: