d3rlpy.q_functions.IQNQFunctionFactory

class d3rlpy.q_functions.IQNQFunctionFactory(n_quantiles=32, embed_size=64)

Implicit Quantile Network Q function factory class.

References

• Dabney et al., Implicit quantile networks for distributional reinforcement learning.

Parameters:

• n_quantiles (int) – the number of quantiles.
• embed_size (int) – the embedding size.

n_quantiles

the number of quantiles.

Type:int

embed_size

the embedding size.

Type:int

Methods

create(encoder, action_size=None)

Returns PyTorch’s Q function module.

Parameters:

• encoder (torch.nn.Module) – an encoder module that processes the observation (and action in continuous action-space) to obtain feature representations.
• action_size (int) – dimension of discrete action-space. If the action-space is continous, None will be passed.

Returns:

Q function object.

Return type:

torch.nn.Module

get_params(deep=False)

Returns Q function parameters.

Returns:Q function parameters. Return type:dict

get_type()

Returns Q function type.

Returns:Q function type. Return type:str

Attributes

TYPE = 'iqn'