update readme, config import

.gitignore

@@ -0,0 +1,2 @@
+*.pyc
+data/

README.md

@@ -111,14 +111,14 @@ optional arguments:
                         what you gave for training). Pass "pretrained" to use
                         pretrained models.
 ```
-Example:
+Run without the `--model_name` argument to test on pre-trained models. Example:
 ```sh
-python sim.py --agent rl-soft --db imdb-M --model_name rl_soft_example.m
+python sim.py --agent rl-soft --db imdb-M
 ```
 
 Hyperparameters
 -------------------------------------------------
-The following agent options can be specified in the config.py file-
+The default hyperparameters for each KB split are in `settings/config_<db_name>.py`. These include:
 1. RL agent options-
   * `nhid`: Number of hidden units
   * `batch`: Batch size

interact.py

@@ -46,6 +46,7 @@
 import numpy as np
 import cPickle as pkl
 import datetime
+import importlib
 
 agent_map = {'rule-no'  :   'nl-rule-no',
         'rl-no'         :   'simple-rl-no',
@@ -77,8 +78,9 @@
 params['dontknow_prob'] = 0.5
 params['sub_prob'] = 0.05
 params['max_first_turn'] = 5
-shutil.copyfile('settings/config_'+params['db']+'.py', 'config.py')
-from config import *
+config = importlib.import_module('settings.config_'+params['db'])
+agent_params = config.agent_params
+dataset_params = config.dataset_params
 for k,v in dataset_params[params['db']].iteritems():
     params[k] = v
 

sim.py

@@ -42,8 +42,9 @@
 params['nlg_slots_path'] = './data/nlg_slot_set.txt'
 params['nlg_model_path'] = './data/pretrained/lstm_tanh_[1470015675.73]_115_120_0.657.p'
 
-shutil.copyfile('settings/config_'+params['db']+'.py', 'config.py')
-from config import *
+config = importlib.import_module('settings.config_'+params['db'])
+agent_params = config.agent_params
+dataset_params = config.dataset_params
 for k,v in dataset_params[params['db']].iteritems():
     params[k] = v
 for k,v in agent_params[agent_map[params['agent_type']]].iteritems():

train.py

@@ -46,8 +46,9 @@
 params['nlg_slots_path'] = './data/nlg_slot_set.txt'
 params['nlg_model_path'] = './data/pretrained/lstm_tanh_[1470015675.73]_115_120_0.657.p'
 
-shutil.copyfile('settings/config_'+params['db']+'.py', 'config.py')
-from config import *
+config = importlib.import_module('settings.config_'+params['db'])
+agent_params = config.agent_params
+dataset_params = config.dataset_params
 for k,v in dataset_params[params['db']].iteritems():
     params[k] = v
 for k,v in agent_params[agent_map[params['agent_type']]].iteritems():