Merge branch 'dev' of github.com:xionghuichen/RLAssistant into dev

README.md

@@ -125,15 +125,18 @@ We build an example project for integrating RLA, which can be seen in ./example/
 We record scalars by `RLA.easy_log.logger`: 
 ```python
 from RLA.easy_log import logger
-
-# scalar variable
-value = 1
-logger.record_tabular("k", value)
-
-# tensorflow summary
 import tensorflow as tf
-summary = tf.Summary()
-logger.log_from_tf_summary(summary)
+from RLA.easy_log.time_step import time_step_holder
+
+for i in range(1000):
+    # time-steps (iterations)
+    time_step_holder.set_time(i)
+    # scalar variable
+    value = 1
+    logger.record_tabular("k", value)
+    # tensorflow summary
+    summary = tf.Summary()
+    logger.log_from_tf_summary(summary)
 
 ```
 **Record checkpoints**
@@ -242,8 +245,9 @@ PS:
 2. An alternative way is building your own NFS for your physical machines and locate data_root to the NFS.
 
 # TODO
-- [ ] video visualization.
+- [ ] support sftp-based sync.
+- [ ] support custom data structure saving and loading.
+- [ ] support video visualization.
 - [ ] add comments and documents to the functions.
 - [ ] add an auto integration script.
 - [ ] download / upload experiment logs through timestamp.
-

RLA/easy_log/exp_loader.py

@@ -32,8 +32,7 @@ class ExperimentLoader(object):
     def __init__(self):
         self.task_name = exp_manager.hyper_param.get('loaded_task_name', None)
         self.load_date = exp_manager.hyper_param.get('loaded_date', None)
-        self.root = getattr(exp_manager, 'root', None)
-        self.data_root = None
+        self.data_root = getattr(exp_manager, 'root', None)
         pass
 
     def config(self, task_name, record_date, root):
@@ -49,12 +48,12 @@ def is_valid_config(self):
             logger.warn("meet invalid loader config when use it")
             logger.warn("load_date", self.load_date)
             logger.warn("task_name", self.task_name)
-            logger.warn("root", self.root)
+            logger.warn("root", self.data_root)
             return False
 
     def import_hyper_parameters(self, hp_to_overwrite: Optional[list] = None):
         if self.is_valid_config:
-            load_tester = Tester.load_tester(self.load_date, self.task_name, self.root)
+            load_tester = Tester.load_tester(self.load_date, self.task_name, self.data_root)
             target_hp = copy.deepcopy(exp_manager.hyper_param)
             target_hp.update(load_tester.hyper_param)
             if hp_to_overwrite is not None:
@@ -75,7 +74,7 @@ def load_from_record_date(self, var_prefix: Optional[str] = None, variable_list:
         :return:
         """
         if self.is_valid_config:
-            loaded_tester = Tester.load_tester(self.load_date, self.task_name, self.root)
+            loaded_tester = Tester.load_tester(self.load_date, self.task_name, self.data_root)
             # load checkpoint
             load_res = {}
             if var_prefix is not None:
@@ -100,7 +99,7 @@ def fork_log_files(self):
         if self.is_valid_config:
             global exp_manager
             assert isinstance(exp_manager, Tester)
-            loaded_tester = Tester.load_tester(self.load_date, self.task_name, self.root)
+            loaded_tester = Tester.load_tester(self.load_date, self.task_name, self.data_root)
             # copy log file
             exp_manager.log_file_copy(loaded_tester)
             # copy attribute
@@ -109,4 +108,4 @@ def fork_log_files(self):
             exp_manager.private_config = loaded_tester.private_config
 
 
-exp_loader = experimental_loader = ExperimentLoader()
+exp_loader = experimental_loader = ExperimentLoader()

RLA/easy_log/logger.py

@@ -688,7 +688,7 @@ def configure(dir=None, format_strs=None, comm=None, framework='tensorflow'):
     if format_strs is None:
         format_strs = os.getenv('OPENAI_LOG_FORMAT', 'stdout,log,csv').split(',')
     format_strs = filter(None, format_strs)
-    output_formats = [make_output_format(f, dir, log_suffix) for f in format_strs]
+    output_formats = [make_output_format(f, dir, log_suffix, framework) for f in format_strs]
     warn_output_formats = make_output_format('warn', dir, log_suffix, framework)
     backup_output_formats = make_output_format('backup', dir, log_suffix, framework)
 

RLA/easy_log/tester.py

@@ -25,7 +25,7 @@
 import shutil
 import argparse
 from typing import Optional, Union, Dict, Any
-from RLA.const import DEFAULT_X_NAME
+from RLA.const import DEFAULT_X_NAME, FRAMEWORK
 import pathspec
 
 def import_hyper_parameters(task_name, record_date):
@@ -187,7 +187,7 @@ def _init_logger(self):
         self.writer = None
         # logger configure
         logger.info("store file %s" % self.pkl_file)
-        logger.configure(self.log_dir, self.private_config["LOG_USED"])
+        logger.configure(self.log_dir, self.private_config["LOG_USED"], framework=self.private_config["DL_FRAMEWORK"])
         for fmt in logger.Logger.CURRENT.output_formats:
             if isinstance(fmt, logger.TensorBoardOutputFormat):
                 self.writer = fmt.writer
@@ -481,7 +481,7 @@ def new_saver(self, max_to_keep, var_prefix=None):
         :param max_to_keep:
         :return:
         """
-        if self.dl_framework == 'tensorflow':
+        if self.dl_framework == FRAMEWORK.tensorflow:
             import tensorflow as tf
             if var_prefix is None:
                 var_prefix = ''
@@ -490,20 +490,20 @@ def new_saver(self, max_to_keep, var_prefix=None):
             for v in var_list:
                 logger.info(v)
             self.saver = tf.train.Saver(var_list=var_list, max_to_keep=max_to_keep, filename=self.checkpoint_dir, save_relative_paths=True)
-        elif self.dl_framework == 'pytorch':
+        elif self.dl_framework == FRAMEWORK.torch:
             self.max_to_keep = max_to_keep
             self.checkpoint_keep_list = []
         else:
             raise NotImplementedError
 
     def save_checkpoint(self, model_dict: Optional[dict]=None, related_variable: Optional[dict]=None):
-        if self.dl_framework == 'tensorflow':
+        if self.dl_framework == FRAMEWORK.tensorflow:
             import tensorflow as tf
             iter = self.time_step_holder.get_time()
             cpt_name = osp.join(self.checkpoint_dir, 'checkpoint')
             logger.info("save checkpoint to ", cpt_name, iter)
             self.saver.save(tf.get_default_session(), cpt_name, global_step=iter)
-        elif self.dl_framework == 'pytorch':
+        elif self.dl_framework == FRAMEWORK.torch:
             import torch
             iter = self.time_step_holder.get_time()
             torch.save(model_dict, f=tester.checkpoint_dir + "checkpoint-{}.pt".format(iter))
@@ -522,7 +522,7 @@ def save_checkpoint(self, model_dict: Optional[dict]=None, related_variable: Opt
         self.add_custom_data(DEFAULT_X_NAME, time_step_holder.get_time(), int, mode='replace')
 
     def load_checkpoint(self):
-        if self.dl_framework == 'tensorflow':
+        if self.dl_framework == FRAMEWORK.tensorflow:
             # TODO: load with variable scope.
             import tensorflow as tf
             cpt_name = osp.join(self.checkpoint_dir)
@@ -532,7 +532,7 @@ def load_checkpoint(self):
             max_iter = ckpt_path.split('-')[-1]
             self.time_step_holder.set_time(max_iter)
             return int(max_iter), None
-        elif self.dl_framework == 'pytorch':
+        elif self.dl_framework == FRAMEWORK.torch:
             import torch
             return self.checkpoint_keep_list[-1], torch.load(tester.checkpoint_dir + "checkpoint-{}.pt".format(self.checkpoint_keep_list[-1]))
 

example/sb3_ppo_example/rla_config.yaml

@@ -18,8 +18,8 @@ LOG_USED:
   - 'tensorboard'
   - 'csv'
 
-# select a DL framework: tensorflow or pytorch.
-DL_FRAMEWORK: 'pytorch'
+# select a DL framework: tensorflow or torch.
+DL_FRAMEWORK: 'torch'
 
 SEND_LOG_FILE: False
 REMOTE_SETTING:

example/sb_ppo_example/rla_config.yaml

@@ -18,7 +18,7 @@ LOG_USED:
   - 'tensorboard'
   - 'csv'
 
-# select a DL framework: tensorflow or pytorch.
+# select a DL framework: tensorflow or torch.
 DL_FRAMEWORK: 'tensorflow'
 
 SEND_LOG_FILE: False

example/simplest_code/rla_config.yaml

@@ -18,7 +18,7 @@ LOG_USED:
   - 'tensorboard'
   - 'csv'
 
-# select a DL framework: tensorflow or pytorch.
+# select a DL framework: "tensorflow" or "torch".
 DL_FRAMEWORK: 'tensorflow'
 
 SEND_LOG_FILE: False