adjust struct of custom_logger and directory

Author: gxywy

README.md

@@ -24,7 +24,7 @@ Add our logger in your code of evaluation
 
 ```python
 from rl_plotter.logger import Logger
-logger = Logger(exp_name="your_exp_name", log_dir, env_name)
+logger = Logger(exp_name="your_exp_name", env_name, seed)
 ····
 logger.update(score=evaluation_score_list, total_steps=current_training_steps)
 ```
@@ -35,6 +35,15 @@ or you can use [OpenAI-baseline](https://github.com/openai/baselines) bench.Moni
 env = logger.monitor_env(env)
 ```
 
+if you want to track other variables:
+
+```python
+custom_logger=logger.new_custom_logger(filename, fieldnames=["variable 1", "variable 2", ..., "variable n"])
+custom_logger.update(fieldvalues=variable_value_list, total_steps=current_training_steps)
+```
+
+
+
 After the training or when you are training your agent, you can plot the learning curves in this way:
 
 - switch to log directory (default: ./)

rl_plotter/logger.py

@@ -36,8 +36,12 @@ def colorize(string, color, bold=False, highlight=False):
 
 class Logger():
     def __init__(self, log_dir="./logs", exp_name=None, env_name=None, seed=0):
+        self.exp_name = exp_name
+        self.env_name = env_name
+        self.seed = seed
+
         num_exps = 0
-        self.log_dir = f"./{log_dir}/{exp_name}_{env_name}_seed{seed}"
+        self.log_dir = f"./{log_dir}/{exp_name.replace('-', '_')}_{env_name.replace('-', '_')}_seed{seed}"
         while True:
             if os.path.exists(f"{self.log_dir}-{str(num_exps)}/"):
                 num_exps += 1
@@ -73,20 +77,15 @@ def update(self, score, total_steps):
         epinfo = {"mean_score": avg_score, "total_steps": total_steps, "std_score": std_score, "max_score": max_score, "min_score": max_score}
         self.logger.writerow(epinfo)
         self.csv_file.flush()
+    
+    def new_custom_logger(self, filename=None, fieldnames=[]):
+        custom_logger = CustomLogger(self.log_dir, self.exp_name, self.env_name, self.seed, filename, fieldnames)
+        return custom_logger
 
 class CustomLogger():
     def __init__(self, log_dir="./logs", exp_name=None, env_name=None, seed=0, filename="logger.csv", fieldnames=[]):
         self.fieldnames = ["total_steps"] + fieldnames
-        num_exps = 0
-        self.log_dir = f"./{log_dir}/{exp_name}_{env_name}_seed{seed}"
-        while True:
-            if os.path.exists(f"{self.log_dir}-{str(num_exps)}/"):
-                num_exps += 1
-            else:
-                self.log_dir += f"-{str(num_exps)}/"
-                os.makedirs(self.log_dir)
-                break
-        self.csv_file = open(self.log_dir + '/' + filename, 'w', encoding='utf8')
+        self.csv_file = open(log_dir + '/' + filename, 'w', encoding='utf8')
         header={"t_start": time.time(), 'env_id' : env_name, 'exp_name': exp_name, 'seed': seed}
         header = '# {} \n'.format(json.dumps(header))
         self.csv_file.write(header)
@@ -95,7 +94,6 @@ def __init__(self, log_dir="./logs", exp_name=None, env_name=None, seed=0, filen
         self.csv_file.flush()
 
     def update(self, fieldvalues, total_steps):
-
         print(colorize(f"\nCustomLogger with fileds: {self.fieldnames}", 'yellow', bold=True))
         print(colorize(f"total_steps: {total_steps}, fieldvalues: {fieldvalues}\n", 'yellow', bold=True))
 

setup.py

@@ -5,7 +5,7 @@
 
 setuptools.setup(
     name="rl_plotter",
-    version="2.2.2",
+    version="2.2.3",
     author="Gong Xiaoyu",
     author_email="[email protected]",
     description="A plotter for reinforcement learning (RL)",