Added success percentage logging and a script for plotting logs

mjrl/algos/batch_reinforce.py

@@ -207,3 +207,8 @@ def log_rollout_statistics(self, paths):
         self.logger.log_kv('stoc_pol_std', std_return)
         self.logger.log_kv('stoc_pol_max', max_return)
         self.logger.log_kv('stoc_pol_min', min_return)
+        try:
+            success_rate = self.env.env.env.evaluate_success(paths)
+            self.logger.log_kv('rollout_success', success_rate)
+        except:
+            pass

mjrl/utils/plot_from_logs.py

@@ -0,0 +1,55 @@
+import os
+import argparse
+import pickle
+import numpy as np
+import matplotlib
+matplotlib.use('Agg')
+import matplotlib.pyplot as plt
+colors = ['b', 'g', 'r', 'c', 'm', 'y', 'k']
+
+parser = argparse.ArgumentParser(description='Script to explore the data generated by an experiment.')
+parser.add_argument('--data', '-d', type=str, required=True, help='location of the .pickle log data file')
+parser.add_argument('--output', '-o', type=str, required=True, help='location to store results as a png')
+parser.add_argument('--xkey', '-x', type=str, default=None, help='the key to use for x axis in plots')
+parser.add_argument('--xscale', '-s', type=int, default=1, help='scaling for the x axis (optional)')
+args = parser.parse_args()
+
+# get inputs and setup output file
+if '.png' in args.output:
+    OUT_FILE = args.output
+else:
+    OUT_FILE = args.output + '/plot.png'
+data = pickle.load(open(args.data, 'rb'))
+xscale = 1 if args.xscale is None else args.xscale
+if args.xkey == 'num_samples':
+    xscale = xscale if 'act_repeat' not in data.keys() else data['act_repeat'][-1]
+
+dict_keys = list(data.keys())
+for k in dict_keys:
+    if len(data[k]) == 1: del(data[k])
+
+# plot layout
+nplt = len(data.keys())
+ncol = 4
+nrow = int(np.ceil(nplt/ncol))
+
+# plot data
+xkey = args.xkey
+start_idx = 2
+end_idx = max([len(data[k]) for k in data.keys()])
+xdata = np.arange(end_idx) if (xkey is None or xkey == 'None') else \
+        [np.sum(data[xkey][:i+1]) * xscale for i in range(len(data[xkey]))]
+
+# make the plot
+plt.figure(figsize=(15,15), dpi=60)
+for idx, key in enumerate(data.keys()):
+    plt.subplot(nrow, ncol, idx+1)
+    plt.tight_layout()
+    try:
+        last_idx = min(end_idx, len(data[key]))
+        plt.plot(xdata[start_idx:last_idx], data[key][start_idx:last_idx], color=colors[idx%7], linewidth=3)
+    except:
+        pass
+    plt.title(key)
+
+plt.savefig(OUT_FILE, dpi=100, bbox_inches="tight")

mjrl/utils/train_agent.py

@@ -114,6 +114,11 @@ def train_agent(job_name, agent,
             mean_pol_perf = np.mean([np.sum(path['rewards']) for path in eval_paths])
             if agent.save_logs:
                 agent.logger.log_kv('eval_score', mean_pol_perf)
+                try:
+                    eval_success = e.env.env.evaluate_success(eval_paths)
+                    agent.logger.log_kv('eval_success', eval_success)
+                except:
+                    pass
 
         if i % save_freq == 0 and i > 0:
             if agent.save_logs: