Merge pull request #11 from pjwozny/vizclean

Vizclean

scripts/evaluate_submission.py

@@ -23,6 +23,7 @@
 import yaml
 import pickle as pkl
 from pathlib import Path
+from visualizeOutputs import construct_stacked_bar_chart
 
 _path = Path(os.path.abspath(__file__))
 
@@ -40,54 +41,6 @@
 import matplotlib.pyplot as plt
 import pandas as pd
 
-def constructStackedBarChart(global_states, 
-                            num_discrete_actions = 10, 
-                            field = "minimum_mitigation_rate_all_regions"):
-
-    """
-    Constructs a Stacked Bar Chart for each timestep of a given run
-    Note: this is only useful where there is a single value per country, 
-        ie self-directed country actions, such as mitigation. 
-
-    Args:
-        global_state(dict): the global state dictionary of a completed run
-        field(str): the name of the field to extract, defualt is minimum_mitigation_rates
-        wandb(wandb-object): an already initialized and open wandb object
-                indicating where the plot should be sent
-    """
-
-
-    rates_over_time = global_states[field]
-    possible_rates = range(0,num_discrete_actions-1)
-    to_plot = {rate:[] for rate in possible_rates}
-
-    #per timestep get rate counts
-    for timestep in range(rates_over_time.shape[0]):
-        current_rates = rates_over_time[timestep,:]
-        current_counter = Counter(current_rates)
-        for rate in possible_rates:
-            #count countries with a particular rate
-            if rate in current_counter.keys():
-                to_plot[rate].append(current_counter[rate])
-            #if no countries have that particular rate
-            else:
-                to_plot[rate].append(0)
-
-    pdf = pd.DataFrame(to_plot)
-    pdf.plot(kind='bar', stacked=True).legend(loc='center left',bbox_to_anchor=(1.0, 0.5))
-    plt.xlabel(f"Countries of a Given {field}")
-    plt.ylabel("Timesteps")
-    plt.title(f"{field} Distribution")
-    plt.show()
-    # wandb.log({f"{field}":plt})
-
-
-
-
-
-
-
-
 # Set logger level e.g., DEBUG, INFO, WARNING, ERROR.
 logging.getLogger().setLevel(logging.ERROR)
 
@@ -335,14 +288,16 @@ def compute_metrics(fetch_episode_states, trainer, framework, submission_file, e
 
         with open("episode_states.pkl", "wb") as f:
             pkl.dump(episode_states[0], f, protocol=pkl.HIGHEST_PROTOCOL)
+
+        if log_config and log_config["enabled"]:
 
-        #log mitigation rate counts of each country over time
-        constructStackedBarChart(episode_states[0],wandb,
-                                    field="mitigation_rate_all_regions")
+            #log mitigation rate counts of each country over time
+            wandb.log({"mitigation_rate Counts Across Time":construct_stacked_bar_chart(episode_states[0],
+                                        field="mitigation_rate_all_regions")})
 
-        #log minimum mitigation rate counts of each country over time
-        constructStackedBarChart(episode_states[0], wandb,
-                                    field="minimum_mitigation_rate_all_regions")
+            #log mitigation rate counts of each country over time
+            wandb.log({"minimum_mitigation_rate Counts Across Time":construct_stacked_bar_chart(episode_states[0],
+                                        field="minimum_mitigation_rate_all_regions")})
 
         for feature in desired_outputs:
             feature_values = [None for _ in range(num_episodes)]