deploy

Visualization Service Deployment

1. How to run

• Quick Model(DEBUG):

  python ./deploy/run.py

• Deployment Model(WSGI Http Server):

  The related parameter usage method can be found in the gunicorn documentation. An example is given below.

  gunicorn -b 0.0.0.0:8888 deploy.run:app --threads 4

2. How to configure

All execution configuration parameters can be set in deploy/dep_config.json. This is a json file, and you need to follow the json writing specifications when editing.

Network Field (net)

• Description

KEY	DES	DEF
port	Backend service open port	not default
app_name	Service access interface path name	not default
session_time_out	The longest life cycle (seconds) in which a session is idle	600

• Example

"net":
{
  "port": 8787,
  "app_name": "convlab2",
  "session_time_out": 300
}

Module Field (nlu, dst, policy, nlg)

The candidate models can be configured under the key values of nlu, dst, policy, nlg. The model under each module needs to set a unique name as the key.

• Description

KEY	DES	DEF
class_path	Target model class relative path	not default
data_set	The data set used by the model	not default
ini_params	The parameters required for the class to be instantiated	{}
model_name	Model name displayed on the front end	model key
max_core	The maximum number of cores this model allows to start	1
preload	If false, this model is not preloaded	true
enable	If false, the system will ignore this configuration	true

• Example

"nlu":
{
  "svm-cam": {
   "class_path": "convlab2.nlu.svm.camrest.nlu.SVMNLU",
   "data_set": "camrest",
   "ini_params": {"mode": "usr"},
   "model_name": "svm-cam",
   "max_core": 1,
   "preload": true,
   "enable": true
 },
 "svm-mul": {
   "class_path": "convlab2.nlu.svm.multiwoz.nlu.SVMNLU",
   "data_set": "multiwoz",
   "ini_params": {"mode": "usr"},
   "model_name": "svm-mul",
   "max_core": 1,
   "preload": false,
   "enable": true
 }
}

3. How to use

After starting up the service, you can open the following url in your web browser:

http://0.0.0.0:[port]/dialog

Then you can choose an available dialog corpus and system configuration. After that, you can start a conversation by input something in the right input area and click [send].

Intermediate Result Modification

In this service, you can restore a dialog turn by modifying the intermediate result which you think is incorrect. For example, if you get an incorrect DST result, you can modify its result in the input area by the left side.

Since you modify the content, a button will appear. Then you can click the button, and the dialog system will restore this turn by masking the DST module and directly use your input as the DST result.