Policy config refactoring (#125)

* Seperate test and train domains

* Add progress bars in ontology embedder

* Update custom_util.py

* Fix custom_util things I broke

* Github master

* Save dialogue ids in prediction file

* Fix bug in ontology enxtraction

* Return dialogue ids in predictions file and fix bugs

* Add setsumbt starting config loader

* Add script to extract golden labels from dataset to match model predictions

* Add more setsumbt configs

* Add option to use local files only in transformers package

* Update starting configurations for setsumbt

* Github master

* Update README.md

* Update README.md

* Update convlab/dialog_agent/agent.py

* Revert custom_util.py

* Update custom_util.py

* Commit unverified chnages :(:(:(:(

* Fix SetSUMBT bug resulting from new torch feature

* Setsumbt bug fixes

* Policy config refactor

* Policy config refactor

* small bug fix in memory with new config path

Co-authored-by: Carel van Niekerk <carel.niekerk@hhu.de>
Co-authored-by: Michael Heck <michael.heck@hhu.de>
Co-authored-by: Christian Geishauser <christian.geishauser@hhu.de>

convlab/policy/ppo/README.md

@@ -11,10 +11,10 @@ If you want to obtain a supervised model for pre-training, please have a look in
 Starting a RL training is as easy as executing
 
 ```sh
-$ python train.py --path=your_environment_config --seed=SEED
+$ python train.py --config_name=your_config_name --seed=SEED
 ```
 
-One example for the environment-config is **semantic_level_config.json**, where parameters for the training are specified, for instance
+One example for the environment-config is **RuleUser-Semantic-RuleDST**, where parameters for the training are specified, for instance
 
 - load_path: provide a path to initialise the model with a pre-trained model, skip the ending .pol.mdl
 - process_num: the number of processes to use during evaluation to speed it up

convlab/policy/ppo/ppo.py

@@ -22,7 +22,7 @@ class PPO(Policy):
 
     def __init__(self, is_train=False, dataset='Multiwoz', seed=0, vectorizer=None):
 
-        with open(os.path.join(os.path.dirname(os.path.abspath(__file__)), 'config.json'), 'r') as f:
+        with open(os.path.join(os.path.dirname(os.path.abspath(__file__)), 'configs' ,'ppo_config.json'), 'r') as f:
             cfg = json.load(f)
         self.save_dir = os.path.join(os.path.dirname(
             os.path.abspath(__file__)), cfg['save_dir'])

convlab/policy/ppo/train.py

@@ -182,8 +182,8 @@ if __name__ == '__main__':
 
     begin_time = datetime.now()
     parser = ArgumentParser()
-    parser.add_argument("--path", type=str, default='convlab/policy/ppo/semantic_level_config.json',
-                        help="Load path for config file")
+    parser.add_argument("--config_name", type=str, default='RuleUser-Semantic-RuleDST',
+                        help="Name of the configuration")
     parser.add_argument("--seed", type=int, default=None,
                         help="Seed for the policy parameter initialization")
     parser.add_argument("--mode", type=str, default='info',
@@ -191,7 +191,8 @@ if __name__ == '__main__':
     parser.add_argument("--save_eval_dials", type=bool, default=False,
                         help="Flag for saving dialogue_info during evaluation")
 
-    path = parser.parse_args().path
+    path = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'configs',
+                        f'{parser.parse_args().config_name}.json')
     seed = parser.parse_args().seed
     mode = parser.parse_args().mode
     save_eval = parser.parse_args().save_eval_dials

convlab/policy/vtrace_DPT/README.md

@@ -31,10 +31,10 @@ We provide several supervised trained models on hugging-face to reproduce the re
 Starting a RL training is as easy as executing
 
 ```sh
-$ python train.py --path=your_environment_config --seed=SEED
+$ python train.py --config_name=your_config_name --seed=SEED
 ```
 
-One example for the environment-config is **semantic_level_config.json**, where parameters for the training are specified, for instance
+One example for the environment-config is **RuleUser-Semantic-RuleDST**, where parameters for the training are specified, for instance
 
 - load_path: provide a path to initialise the model with a pre-trained model, skip the ending .pol.mdl
 - process_num: the number of processes to use during evaluation to speed it up

convlab/policy/vtrace_DPT/memory.py

@@ -17,7 +17,9 @@ class Memory:
 
     def __init__(self, seed=0):
 
-        with open(os.path.join(os.path.dirname(os.path.abspath(__file__)), 'config.json'), 'r') as f:
+        dir_name = os.path.dirname(os.path.abspath(__file__))
+        self.config_path = os.path.join(dir_name, 'configs', 'multiwoz21_dpt.json')
+        with open(self.config_path, 'r') as f:
             cfg = json.load(f)
 
         self.batch_size = cfg.get('batchsz', 32)

convlab/policy/vtrace_DPT/train.py

@@ -101,8 +101,8 @@ if __name__ == '__main__':
 
     begin_time = datetime.now()
     parser = ArgumentParser()
-    parser.add_argument("--path", type=str, default='convlab/policy/vtrace_DPT/semantic_level_config.json',
-                        help="Load path for config file")
+    parser.add_argument("--config_name", type=str, default='RuleUser-Semantic-RuleDST',
+                        help="Name of the configuration")
     parser.add_argument("--seed", type=int, default=None,
                         help="Seed for the policy parameter initialization")
     parser.add_argument("--mode", type=str, default='info',
@@ -110,7 +110,8 @@ if __name__ == '__main__':
     parser.add_argument("--save_eval_dials", type=bool, default=False,
                         help="Flag for saving dialogue_info during evaluation")
 
-    path = parser.parse_args().path
+    path = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'configs',
+                        f"{parser.parse_args().config_name}.json")
     seed = parser.parse_args().seed
     mode = parser.parse_args().mode
     save_eval = parser.parse_args().save_eval_dials

convlab/policy/vtrace_DPT/vtrace.py

@@ -26,7 +26,7 @@ class VTRACE(nn.Module, Policy):
         super(VTRACE, self).__init__()
 
         dir_name = os.path.dirname(os.path.abspath(__file__))
-        self.config_path = os.path.join(dir_name, 'config.json')
+        self.config_path = os.path.join(dir_name, 'configs', 'multiwoz21_dpt.json')
 
         with open(self.config_path, 'r') as f:
             cfg = json.load(f)