metaphor.api_client

Low level client interface for Metaphor server.

This module expose procedures that call directly the metaphor server through the RESTful interface implemented with Sanic.

meta_client.api_client.add_user(user_ame='')

Add a user to the registered user list.

parameter:

• user_ame -> name of the user to add

meta_client.api_client.check_user(username='', getID=False)

Check the user situation.

return :

user registered	no	yes
getID True	-1	ID
getID False	warning	3-tuple : name, ID, belongings dictionary

meta_client.api_client.contact()

Establish a first contact with the metaphor server. It returns a welcome message.

parameters:

• application (optional) -> the application name get in contact with. Defaul metaphor

meta_client.api_client.create_model(userid, data='', raw=False)

create a new model.

• data -> dictionary describing creation parameters.

  • training -> boolean giving the purpose of creation

  • source -> data file or data file content

  • modelname -> name imposed to the model

  • trainable -> training capacity (default True)

  • inputs -> input number (default 1)

  • outputs -> output number (default 1)

  • hidden -> number of hidden neurons (default 2)

  • activfunc -> activation fonction (default TANH)

  • statelayer -> state layer (default -1 )

  • statenodeindex index of state node (default -1)

  • order order of the network default 0)

  • nosynapse -> build model without any synapses (default False)

  • existingnode -> already existing nodes (default 0)

  • polytype -> type de polynome (default 0)

  • classif -> classifier (deafult False)

  • verbose=0,

  • data=None,

  • dataformat=None,

  • normalize=False,

  • weights=None,

  • disp=0,

  • callback=None unused

• raw -> If True, return raw request

meta_client.api_client.delete_models(userid, modelid)

Delete a registered model.

parameters:

• userid -> user ID

• modelid -> model ID belonging to the designed user. modelid may be “*” to delete all the user’s models.

meta_client.api_client.delete_user(user_name='')

Unregister the designed user.

All the user belongings are destroyed.

parameter:

• user_name -> name of the user to unregister. If this name is “*”, all the registered users are unregistered.

meta_client.api_client.get_function(userid, modelid, root_func_name, data='')

Call a function to apply to the designed model. The function called may not modify the sate of the designed model.

parameters:

• userid -> user ID

• modelid -> model ID belonging to the designed user

• root_func_name, data -> function to call. data is related to function:

  function name	data
  paramCount	None
  dataCount	None
  newWeights	{‘stddev’: stddev, ‘bias0’: bias0}
  getTrainingData	{‘prefix’: None}
  inputNames	None
  outputNames	None
  inputs	None
  outputs	None
  inputNorm	None
  outputNorm	None
  targets	None
  getWeights	{‘prefix’: None}

meta_client.api_client.get_model_attribute(userid, modelid, data='')

Retrieve model attributes.

parameters:

• userid -> user ID

• modelid -> model ID belonging to the designed user.

• data -> dictionary describing the attribute to read.

  • {‘attributes’: <attribute name>}

return depending on the attribute.

meta_client.api_client.get_model_list(userid)

Retrieve the list of models belongings to the given user.

parameter:

• userid -> ID of the user to examine.

meta_client.api_client.get_user_name(user_ID)

Retrieve user name.

parameter:

• user_ID -> ID to find

return user name if found, else empty string.

meta_client.api_client.get_users()

Retrieve registered user dictionary.

• keys -> user ID(s)

• values -> user names

meta_client.api_client.initiate_connection(server='0.0.0.0', port=5005, application='metaphor', APIver=1.0)

Initiate metaphor client interface routes and establish contact.

parameters:

• server -> IP address or name of the server machine. default “0.0.0.0”

• port -> Number of port. default 5005

• application -> Application name. default ‘metaphor’

• APIver -> Version of the interface. default APIversion

return the server welcome message if success, otherwise an error message.

meta_client.api_client.load_model(userid, data='', getID=False, raw=False)

load a new model from file.

• datas :

  • source -> file or file content

  • ‘modelname’ -> name imposed to the model

  • ‘descriptor’ -> smiles code or similar descriptor

  • ‘docompile’ ->

raw -> If True, return raw request

meta_client.api_client.post_file(userid, filepath=None, dest='', extra={})

Upload a file to the server.

parameters:

• filepath -> path of the file to upload

• dest -> relative destination path in the server

• extra -> unused

meta_client.api_client.post_function(userid, modelid, root_func_name, data='')

Call a function to apply to the designed model. The function called may modify the sate of the designed model.

parameters:

• userid -> user ID

• modelid -> model ID belonging to the designed user

• root_func_name -> function to call. Availables functions are:

  • transfer

  • train

  • reverse_train

  • setWeights

• data -> a dictionary depending upon the called function:

  • transfer:

    • inputs -> input vector.

    • index -> index of training data to use as inputs.

    • weights -> parameters to use. If None, current parameters are used.

  • train ->

    • startWeights -> starting parameter set to strat training with.

    • epochs -> maximum number of epochs allowed to training job

    • trainstyle -> unused

    • callback -> unused

  • reverse_train ->

    • outputtarget -> output value to target

    • inputs -> starting inputs

    • fixedinputs -> vector of boolean giving the input mobility freedom

    • epochs -> maximum number of epochs allowed to training job

  • setWeights ->

    • weights -> parameter vector to affect

    • prefix -> None

meta_client.api_client.set_model_attribute(userid, modelid, data='')

Write an attribute to the model. Can be used for user defined attributes.

parameters:

• userid -> user ID

• modelid -> model ID belonging to the designed user

• data -> dictionary describing the attribute to write

  • {‘attributes’: <attribute name>}

metaphor.client_toolbox

This module defines the routes for metaphor server.

meta_client.client_toolbox.initiate_routes(server='0.0.0.0', port=5005, appli='metaphor', APIver=1.0)

Initiate metaphor client interface routes.

parameters:

• server -> IP address or name of the server machine. default “0.0.0.0”

• port -> Number of port. default 5005

• application -> Application name. default ‘metaphor’

• APIversion -> Version of the interface. default APIversion