Module fontai.config.deployment

Classes

class Config (**data: Any)

Base class for ML pipelane stage configuration objects

Args

input_path : str, optional

object to list and retrieve input files to be processed

output_path : str, optional

object to persist output bojects

yaml : yml.YAML

parsed YAML from supplied configuration file

Create a new model by parsing and validating input data from keyword arguments.

Raises ValidationError if the input data cannot be parsed to form a valid model.

Expand source code

class Config(BasePipelineTransformConfig):

  model: tf.keras.Model
  sampler: t.Callable
  charset_size: PositiveInt
  grid: Grid
  dash_args: t.Dict
  plotter: plotters.AlphabetPlotter

Ancestors

• BasePipelineTransformConfig
• pydantic.main.BaseModel
• pydantic.utils.Representation

Class variables

var charset_size : pydantic.types.PositiveInt

var dash_args : Dict[~KT, ~VT]

var grid : Grid

var model : tensorflow.python.keras.engine.training.Model

var plotter : AlphabetPlotter

var sampler : Callable

class ConfigHandler

Wrapper for deployment's configuration logic.

Expand source code

class ConfigHandler(BaseConfigHandler):
  """
  Wrapper for deployment's configuration logic.
  
  """
  @classmethod
  def get_config_schema(self):
    """
    YAML configuration schema:
    
    decoder_path: Path to generative model that takes point in style space and one-hot-encoded character representations and output an image
    charset_size: Dimensionality of the characters' one-hot-encoded representation
    style_vector_size: Dimensionality of style space
    style_sampler (optional, defaults to 'normal'): distribution from numpy.random to sample from style space. 'size' should be its only argument
    min_style_value (optional, defaults to -3): minimum componentwise style value in style space grid
    max_style_value (optional, defaults to 3): maximum componentwise style value in style space grid
    style_grid_size (optional, defaults to 100): Componentwise number of bins in style grid


    """
    
    schema = yml.Map({
      "decoder_path" : self.IO_CONFIG_SCHEMA,
      "charset_size": yml.Int(),
      "style_vector_size": yml.Int(),
      yml.Optional("style_sampler", default="normal"): yml.Str(),
      yml.Optional("min_style_value", default=-3): yml.Float(),
      yml.Optional("max_style_value", default=3): yml.Float(),
      yml.Optional("style_grid_size", default=100): yml.Int(),
      yml.Optional("plotter", default="SAAEAlphabetPlotter"): yml.Str(),
      yml.Optional("dash_args", default={}): yml.MapPattern(
        yml.Str(), 
        self.yaml_to_obj.ANY_PRIMITIVES) | yml.EmptyDict()
    })

    return schema

  def instantiate_config(self, config: yml.YAML) -> Config:
    """
    Processes a YAML instance to produce an Config instance.
        
    Args:
        config (yml.YAML): YAML object from the strictyaml library
    
    Returns:
        Config: Configuration object
    
    """
    

    
    model = tf.keras.models.load_model(config.get("decoder_path").text)
    sampler = getattr(rnd, config.get("style_sampler").text)
    grid = Grid(
      dim = config.get("style_vector_size").data,
      lowest = config.get("min_style_value").data,
      largest = config.get("max_style_value").data,
      size = config.get("style_grid_size").data)

    dash_args = config.get("dash_args").data

    plotter = getattr(plotters,config.get("plotter").text)()

    return Config(
      model = model,
      dash_args = dash_args,
      sampler = sampler,
      plotter = plotter,
      charset_size = config.get("charset_size").data,
      grid = grid,
      yaml = config)

  def other_setup(self):
    self.model_factory = ModelFactory()

Ancestors

• BaseConfigHandler
• abc.ABC

Static methods

def get_config_schema()

YAML configuration schema:

decoder_path: Path to generative model that takes point in style space and one-hot-encoded character representations and output an image charset_size: Dimensionality of the characters' one-hot-encoded representation style_vector_size: Dimensionality of style space style_sampler (optional, defaults to 'normal'): distribution from numpy.random to sample from style space. 'size' should be its only argument min_style_value (optional, defaults to -3): minimum componentwise style value in style space grid max_style_value (optional, defaults to 3): maximum componentwise style value in style space grid style_grid_size (optional, defaults to 100): Componentwise number of bins in style grid

Expand source code

@classmethod
def get_config_schema(self):
  """
  YAML configuration schema:
  
  decoder_path: Path to generative model that takes point in style space and one-hot-encoded character representations and output an image
  charset_size: Dimensionality of the characters' one-hot-encoded representation
  style_vector_size: Dimensionality of style space
  style_sampler (optional, defaults to 'normal'): distribution from numpy.random to sample from style space. 'size' should be its only argument
  min_style_value (optional, defaults to -3): minimum componentwise style value in style space grid
  max_style_value (optional, defaults to 3): maximum componentwise style value in style space grid
  style_grid_size (optional, defaults to 100): Componentwise number of bins in style grid


  """
  
  schema = yml.Map({
    "decoder_path" : self.IO_CONFIG_SCHEMA,
    "charset_size": yml.Int(),
    "style_vector_size": yml.Int(),
    yml.Optional("style_sampler", default="normal"): yml.Str(),
    yml.Optional("min_style_value", default=-3): yml.Float(),
    yml.Optional("max_style_value", default=3): yml.Float(),
    yml.Optional("style_grid_size", default=100): yml.Int(),
    yml.Optional("plotter", default="SAAEAlphabetPlotter"): yml.Str(),
    yml.Optional("dash_args", default={}): yml.MapPattern(
      yml.Str(), 
      self.yaml_to_obj.ANY_PRIMITIVES) | yml.EmptyDict()
  })

  return schema

Methods

def instantiate_config(self, config: strictyaml.representation.YAML) ‑> Config

Processes a YAML instance to produce an Config instance.

Args

config : yml.YAML

YAML object from the strictyaml library

Returns

Config

Configuration object

Expand source code

def instantiate_config(self, config: yml.YAML) -> Config:
  """
  Processes a YAML instance to produce an Config instance.
      
  Args:
      config (yml.YAML): YAML object from the strictyaml library
  
  Returns:
      Config: Configuration object
  
  """
  

  
  model = tf.keras.models.load_model(config.get("decoder_path").text)
  sampler = getattr(rnd, config.get("style_sampler").text)
  grid = Grid(
    dim = config.get("style_vector_size").data,
    lowest = config.get("min_style_value").data,
    largest = config.get("max_style_value").data,
    size = config.get("style_grid_size").data)

  dash_args = config.get("dash_args").data

  plotter = getattr(plotters,config.get("plotter").text)()

  return Config(
    model = model,
    dash_args = dash_args,
    sampler = sampler,
    plotter = plotter,
    charset_size = config.get("charset_size").data,
    grid = grid,
    yaml = config)

def other_setup(self)

Expand source code

def other_setup(self):
  self.model_factory = ModelFactory()

Inherited members

• BaseConfigHandler:
  • from_file
  • from_string

class Grid (**data: Any)

Represents a grid in style space that can be explored through an interactive Dash app

Create a new model by parsing and validating input data from keyword arguments.

Raises ValidationError if the input data cannot be parsed to form a valid model.

Expand source code

class Grid(BaseModel):

  """Represents a grid in style space that can be explored through an interactive Dash app
  """

  dim: PositiveInt #dimensionality in style space
  lowest: float #lowest grid value componentwise
  largest: float #largest grid value componentwise
  size: PositiveInt #componentwise grid zie

Ancestors

• pydantic.main.BaseModel
• pydantic.utils.Representation

Class variables

var dim : pydantic.types.PositiveInt

var largest : float

var lowest : float

var size : pydantic.types.PositiveInt