diff --git a/baselines/run_baseline_parallel.py b/baselines/run_baseline_parallel.py
index 364c5afe5..3960687e6 100644
--- a/baselines/run_baseline_parallel.py
+++ b/baselines/run_baseline_parallel.py
@@ -8,6 +8,7 @@
 from stable_baselines3.common.utils import set_random_seed
 from stable_baselines3.common.callbacks import CheckpointCallback
 from argparse_pokemon import *
+import os
 
 def make_env(rank, env_conf, seed=0):
     """
@@ -39,8 +40,10 @@ def _init():
             }
     
     env_config = change_env(env_config, args)
-    
-    num_cpu = 44 #64 #46  # Also sets the number of episodes per training iteration
+
+    # Set the number of cpus dynamically
+    num_cpu = os.cpu_count()
+
     env = SubprocVecEnv([make_env(i, env_config) for i in range(num_cpu)])
     
     checkpoint_callback = CheckpointCallback(save_freq=ep_length, save_path=sess_path,
@@ -62,4 +65,4 @@ def _init():
         model = PPO('CnnPolicy', env, verbose=1, n_steps=ep_length, batch_size=512, n_epochs=1, gamma=0.999)
     
     for i in range(learn_steps):
-        model.learn(total_timesteps=(ep_length)*num_cpu*1000, callback=checkpoint_callback)
\ No newline at end of file
+        model.learn(total_timesteps=(ep_length)*num_cpu*1000, callback=checkpoint_callback)