tatk.policy.vhus package

Submodules

tatk.policy.vhus.train module

@author: truthless

class tatk.policy.vhus.train.VHUS_Trainer(config, manager, goal_gen)

Bases: object

__init__(config, manager, goal_gen)

Initialize self. See help(type(self)) for accurate signature.

imit_test(epoch, best)

provide an unbiased evaluation of the user simulator fit on the training dataset

imitating(epoch)

train the user simulator by simple imitation learning (behavioral cloning)

save(directory, epoch)
test()
user_loop(data)
tatk.policy.vhus.train.batch_iter(x, y, z, batch_size=64)

tatk.policy.vhus.usermodule module

@author: truthless

class tatk.policy.vhus.usermodule.Decoder(vocab_size, max_len, embed_size, hidden_size, sos_id=2, eos_id=3, n_layers=1, rnn_cell='GRU', input_dropout_p=0, dropout_p=0, use_attention=False)

Bases: torch.nn.modules.module.Module

KEY_ATTN_SCORE = 'attention_score'
KEY_LENGTH = 'length'
KEY_SEQUENCE = 'sequence'
__init__(vocab_size, max_len, embed_size, hidden_size, sos_id=2, eos_id=3, n_layers=1, rnn_cell='GRU', input_dropout_p=0, dropout_p=0, use_attention=False)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(inputs=None, encoder_hidden=None, encoder_outputs=None, function=<built-in method log_softmax of type object>, teacher_forcing_ratio=0)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

forward_step(input_var, hidden, encoder_outputs, function)
class tatk.policy.vhus.usermodule.Encoder(vocab_size, embed_size, hidden_size, input_dropout_p=0, dropout_p=0, n_layers=1, rnn_cell='GRU', variable_lengths=False, embedding=None, update_embedding=True)

Bases: torch.nn.modules.module.Module

__init__(vocab_size, embed_size, hidden_size, input_dropout_p=0, dropout_p=0, n_layers=1, rnn_cell='GRU', variable_lengths=False, embedding=None, update_embedding=True)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(input_var, input_lengths=None)

Applies a multi-layer RNN to an input sequence. Args:

input_var (batch, seq_len): tensor containing the features of the input sequence. input_lengths (list of int, optional): A list that contains the lengths of sequences

in the mini-batch

Returns: output, hidden

  • output (batch, seq_len, hidden_size): variable containing the encoded features of the input sequence

  • hidden (num_layers * num_directions, batch, hidden_size): variable containing the features in the hidden state h

class tatk.policy.vhus.usermodule.VHUS(cfg, voc_goal_size, voc_usr_size, voc_sys_size)

Bases: torch.nn.modules.module.Module

__init__(cfg, voc_goal_size, voc_usr_size, voc_sys_size)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

forward(goals, goals_length, posts, posts_length, origin_responses=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

select_action(goal, goal_length, post, post_length)
Parameters

  • goal – [goal_len]

  • goal_length – []

  • post – [sen_len, word_len]

  • post_length – [sen_len]

Returns

[act_len], [1]

tatk.policy.vhus.usermodule.batch_gather_3_1(inputs, dim)
Args:

inputs (batchsz, sen_len, embed_dim) dim (batchsz)

Returns:

output (batch, embed_dim)

tatk.policy.vhus.usermodule.batch_gather_4_2(inputs, dim)
Args:

inputs (batchsz, sen_len, word_len, embed_dim) dim (batchsz, sen_len)

Returns:

output (batch, sen_len, embed_dim)

tatk.policy.vhus.usermodule.reparameterize(mu, logvar)

tatk.policy.vhus.util module

tatk.policy.vhus.util.capital(da)
tatk.policy.vhus.util.kl_gaussian(argu)
tatk.policy.vhus.util.padding(old, l)

pad a list of different lens “old” to the same len “l”

tatk.policy.vhus.util.padding_data(data)

tatk.policy.vhus.vhus module

class tatk.policy.vhus.vhus.UserPolicyVHUSAbstract(archive_file, model_file)

Bases: tatk.policy.policy.Policy

__init__(archive_file, model_file)

Initialize self. See help(type(self)) for accurate signature.

get_goal()
init_session()

Init the class variables for a new session.

is_terminated()
load(archive_file, model_file, filename)
predict(state)

Predict an user act based on state and preorder system action.

Args:
state (tuple):

Dialog state.

Returns:
usr_action (tuple):

User act.

session_over (boolean):

True to terminate session, otherwise session continues.