Add graphs

encoder.py

@@ -109,10 +109,7 @@ class PixelEncoder(nn.Module):
         out_dim = OUT_DIM[num_layers]
         self.fc = nn.Linear(num_filters * out_dim * out_dim, self.feature_dim * 2)
         self.ln = nn.LayerNorm(self.feature_dim * 2)
-<<<<<<< HEAD
         self.combine = nn.Linear(self.feature_dim + 6, self.feature_dim)
-=======
->>>>>>> origin/tester_1
 
         self.outputs = dict()
 
@@ -157,11 +154,7 @@ class PixelEncoder(nn.Module):
 
         out = self.reparameterize(mu, logstd)
         self.outputs['tanh'] = out
-<<<<<<< HEAD
         return out, mu, logstd
-=======
-        return out
->>>>>>> origin/tester_1
 
     def copy_conv_weights_from(self, source):
         """Tie convolutional layers"""

graphs_plot.py

@@ -1,10 +1,11 @@
 import os
+import numpy as np
 import pandas as pd
 import seaborn as sns
 import matplotlib.pyplot as plt
 from tensorboard.backend.event_processing.event_accumulator import EventAccumulator
 
-
+"""
 def tabulate_events(dpath):
     files = os.listdir(dpath)[0]
     summary_iterators = [EventAccumulator(os.path.join(dpath, files)).Reload()]
@@ -43,7 +44,43 @@ for tag, values in events.items():
 
 df = pd.DataFrame(data)
 print(df.head())
+exit()
 
 plt.figure(figsize=(10,6))
 sns.lineplot(data=df, x='step', y='value', hue='tag', ci='sd')
 plt.show()
+"""
+
+
+from tensorboard.backend.event_processing import event_accumulator
+
+
+
+def data_from_tb(files):
+    all_steps, all_rewards = [], []
+    for file in files:
+        ea = event_accumulator.EventAccumulator(file, size_guidance={'scalars': 0})
+        ea.Reload()
+
+        episode_rewards = ea.Scalars('train/episode_reward')
+        steps = [event.step for event in episode_rewards][:990000]
+        rewards = [event.value for event in episode_rewards][:990000]
+        all_steps.append(steps)
+        all_rewards.append(rewards)
+    return all_steps, all_rewards
+
+
+files = ['/home/vedant/pytorch_sac_ae/log/runs/tb_21_05_2023-13_19_36/events.out.tfevents.1684667976.cpswkstn6-nvidia4090.1749060.0',
+         '/home/vedant/pytorch_sac_ae/log/runs/tb_22_05_2023-09_56_30/events.out.tfevents.1684742190.cpswkstn6-nvidia4090.1976229.0']
+
+all_steps, all_rewards = data_from_tb(files)
+mean_rewards = np.mean(all_rewards, axis=0)
+std_rewards = np.std(all_rewards, axis=0)
+mean_steps = np.mean(all_steps, axis=0) 
+
+df = pd.DataFrame({'Steps': mean_steps,'Rewards': mean_rewards,'Standard Deviation': std_rewards})
+
+sns.relplot(x='Steps', y='Rewards', kind='line', data=df, ci="sd")
+plt.fill_between(df['Steps'], df['Rewards'] - df['Standard Deviation'], df['Rewards'] + df['Standard Deviation'], color='b', alpha=.1)
+plt.title("Mean Rewards vs Steps with Standard Deviation")
+plt.show()

sac_ae.py

@@ -417,14 +417,13 @@ class SacAeAgent(object):
         h_dist_pred = torch.distributions.Normal(mean, std)
         enc_loss = torch.distributions.kl.kl_divergence(h_dist_enc, h_dist_pred).mean() * 1e-2
         
-        """
         with torch.no_grad():
             z_pos, _ , _ = self.critic_target.encoder(next_obs_list[-1])
             z_out = self.critic_target.encoder.combine(torch.concat((z_pos, action), dim=-1))
         logits = self.lb_loss.compute_logits(h, z_out)
         labels = torch.arange(logits.shape[0]).long().to(self.device)
         lb_loss = nn.CrossEntropyLoss()(logits, labels) * 1e-2
-        """
+        
         #with torch.no_grad():
         #    z_pos, _ , _ = self.critic.encoder(next_obs_list[-1])
         #ub_loss = club_loss(state_enc["sample"], mean, state_enc["logvar"], h) * 1e-1
@@ -437,7 +436,7 @@ class SacAeAgent(object):
 
         ub_loss = torch.tensor(0.0)
         #enc_loss = torch.tensor(0.0)
-        lb_loss = torch.tensor(0.0)
+        #lb_loss = torch.tensor(0.0)
         #rec_loss = torch.tensor(0.0)
         loss = rec_loss + enc_loss + lb_loss + ub_loss
         self.encoder_optimizer.zero_grad()

train.py

@@ -28,10 +28,7 @@ def parse_args():
     parser.add_argument('--frame_stack', default=3, type=int)
     parser.add_argument('--img_source', default=None, type=str, choices=['color', 'noise', 'images', 'video', 'none'])
     parser.add_argument('--resource_files', type=str)
-<<<<<<< HEAD
     parser.add_argument('--resource_files_test', type=str)
-=======
->>>>>>> origin/tester_1
     parser.add_argument('--total_frames', default=10000, type=int)
     # replay buffer
     parser.add_argument('--replay_buffer_capacity', default=100000, type=int)