parlai.core.loader¶

Functions for loading modules for agents, tasks and teachers, worlds, and scripts.

These functions are largely for converting strings specified in opts (like for –task) to the appropriate module.

class parlai.core.loader.script_registration(klass, hidden, aliases)¶

Bases: tuple

aliases¶: Alias for field number 2

hidden¶: Alias for field number 1

klass¶: Alias for field number 0

parlai.core.loader.register_agent(name: str) → Callable[[Type], Type][source]¶

Register an agent to be available in command line calls.

>>> @register_agent("my_agent")
... class MyAgent:
...     pass

parlai.core.loader.register_script(name: str, aliases=None, hidden=False)[source]¶

Register an agent to be available in command line calls.

>>> @register_script("my_script")
... class MyScript:
...     pass

parlai.core.loader.register_teacher(name: str) → Callable[[Type], Type][source]¶

Register a teacher to be available as a command line.

>>> @register_teacher("my_teacher")
... class MyTeacher:
...    pass

parlai.core.loader.load_agent_module(agent_path: str)[source]¶

Return the module for an agent specified by --model.

Can be formatted in several different ways:

full: -m parlai.agents.seq2seq.seq2seq:Seq2seqAgent
shorthand: -m seq2seq, which will check both paths parlai.agents.seq2seq.seq2seq:Seq2seqAgent and parlai.agents.seq2seq.agents:Seq2seqAgent
half-shorthand: -m seq2seq/variant, which will check the path parlai.agents.seq2seq.variant:VariantAgent

The base path to search when using shorthand formats can be changed from “parlai” to “parlai_internal” by prepending “internal:” to the path, e.g. “internal:seq2seq”.

To use agents in projects, you can prepend “projects:” and the name of the project folder to model arguments, e.g. “projects:personachat:kvmemnn” will translate to projects/personachat/kvmemnn.

Parameters: agent_path – path to model class in one of the above formats.
Returns: module of agent

parlai.core.loader.load_task_module(taskname: str)[source]¶

Get the module containing all teacher agents for the task specified by –task.

Parameters: taskname – path to task class in one of the formats

The valid formats for taskname are:

full: -t parlai.tasks.babi.agents:DefaultTeacher

shorthand: -t babi, which will check parlai.tasks.babi.agents:DefaultTeacher

shorthand specific: -t babi:task10k, which will check parlai.tasks.babi.agents:Task10kTeacher

Returns: module containing all teacher agents for a task

parlai.core.loader.load_teacher_module(taskname: str)[source]¶

Get the module of the teacher agent specified by –task.

Can be formatted in several different ways:

full: -t parlai.tasks.babi.agents:DefaultTeacher
shorthand: -t babi, which will check parlai.tasks.babi.agents:DefaultTeacher
shorthand specific: -t babi:task10k, which will check parlai.tasks.babi.agents:Task10kTeacher

The base path to search when using shorthand formats can be changed from “parlai” to “parlai_internal” by prepending “internal:” to the path, e.g. “internal:babi”.

Options can be sent to the teacher by adding an additional colon, for example -t babi:task10k:1 directs the babi Task10kTeacher to use task number 1.

Parameters: taskname – path to task class in one of the above formats.
Returns: teacher module

parlai.core.loader.load_world_module(taskname: str, interactive_task: bool = False, selfchat_task: bool = False, num_agents: Optional[int] = None, default_world=None)[source]¶

Load the world module for the specific environment. If not enough information is to determine which world should be loaded, returns None.

Parameters

taskname – path to task class in one of the above formats
interactive_task – whether or not the task is interactive
num_agents – number of agents in the world; this may not be known a priori
default_world – default world to return if specified

Returns

World module (or None, if not enough info to determine is present)