dymad.models.components

Module Attributes

ENC_MAP	Mapping of encoder names to encoder functions.
DEC_MAP	Mapping of decoder names to decoder functions.
FZU_MAP	Mapping of feature concatenation names to functions.
DYN_MAP	Mapping of dynamics composer names to functions.
LIN_MAP	Mapping of linear evaluation and features functions.

Functions

dec_auto(net, z, w)	Generic decoder function.
dec_graph(net, z, w)	Graph decoder function.
dec_iden(net, z, w)	Identity decoder function.
dec_node(net, z, w)	Node-wise decoder function.
dyn_direct(net, s, z, w)	Processing without control inputs.
dyn_graph_direct(net, s, z, w)	Processing by GNN.
dyn_graph_skip(net, s, z, w)	Processing by GNN with skip connection.
dyn_skip(net, s, z, w)	Processing with skip connection.
enc_graph_auto(net, w)	Using GNN in EncAuto.
enc_graph_ctrl(net, w)	Using GNN in EncCtrl.
enc_graph_iden(net, w)	Identity encoder function for graph data.
enc_iden(net, w)	Identity encoder function.
enc_node_auto(net, w)	Using EncAuto for each node of graph.
enc_node_ctrl(net, w)	Using EncCtrl for each node of graph.
enc_node_raw_ctrl(net, w)	Using EncCtrl for each node of graph, letting encoder handle the concatenation.
enc_raw_ctrl(net, w)	Let encoder handle states and controls.
enc_smpl_auto(net, w)	Only encodes states.
enc_smpl_ctrl(net, w)	Encodes states and controls.
linear_eval_graph(mdl, w)	Compute linear evaluation, dz, and states, z, for the model.
linear_eval_smpl(mdl, w)	Compute linear evaluation, dz, and states, z, for the model.
linear_features_graph(mdl, w)	Compute linear features, f, and outputs, dz, for the model.
linear_features_smpl(mdl, w)	Compute linear features, f, and outputs, dz, for the model.
zu_blin_no_const(z, w)	Compute bilinear features without constant term.
zu_blin_no_const_graph(z, w)	Compute bilinear features without constant term for graph data.
zu_blin_with_const(z, w)	Compute bilinear features with constant term.
zu_blin_with_const_graph(z, w)	Compute bilinear features with constant term for graph data.
zu_cat_none(z, w)	No concatenation, just return z.
zu_cat_smpl(z, w)	Simple concatenation of z and u.
zu_cat_smpl_graph(z, w)	Simple concatenation of z and u on graph.

dymad.models.components.DEC_MAP = {'auto': <function dec_auto>, 'graph': <function dec_graph>, 'iden': <function dec_iden>, 'node': <function dec_node>}

Mapping of decoder names to decoder functions.

dymad.models.components.DYN_MAP = {'direct': <function dyn_direct>, 'graph_direct': <function dyn_graph_direct>, 'graph_skip': <function dyn_graph_skip>, 'skip': <function dyn_skip>}

Mapping of dynamics composer names to functions.

dymad.models.components.ENC_MAP = {'graph_auto': <function enc_graph_auto>, 'graph_ctrl': <function enc_graph_ctrl>, 'graph_iden': <function enc_graph_iden>, 'iden': <function enc_iden>, 'node_auto': <function enc_node_auto>, 'node_ctrl': <function enc_node_ctrl>, 'node_iden': <function enc_iden>, 'node_raw_auto': <function enc_node_auto>, 'node_raw_ctrl': <function enc_node_raw_ctrl>, 'raw_auto': <function enc_smpl_auto>, 'raw_ctrl': <function enc_raw_ctrl>, 'smpl_auto': <function enc_smpl_auto>, 'smpl_ctrl': <function enc_smpl_ctrl>}

Mapping of encoder names to encoder functions.

dymad.models.components.FZU_MAP = {'blin_no_const': <function zu_blin_no_const>, 'blin_with_const': <function zu_blin_with_const>, 'cat': <function zu_cat_smpl>, 'graph_blin_no_const': <function zu_blin_no_const_graph>, 'graph_blin_with_const': <function zu_blin_with_const_graph>, 'graph_cat': <function zu_cat_smpl_graph>, 'none': <function zu_cat_none>}

Mapping of feature concatenation names to functions.

dymad.models.components.LIN_MAP = {'graph': (<function linear_eval_graph>, <function linear_features_graph>), 'smpl': (<function linear_eval_smpl>, <function linear_features_smpl>)}

Mapping of linear evaluation and features functions.

dymad.models.components.dec_auto(net, z, w)

Generic decoder function.

Return type:

Tensor

dymad.models.components.dec_graph(net, z, w)

Graph decoder function.

Return type:

Tensor

dymad.models.components.dec_iden(net, z, w)

Identity decoder function.

Return type:

Tensor

dymad.models.components.dec_node(net, z, w)

Node-wise decoder function.

Return type:

Tensor

dymad.models.components.dyn_direct(net, s, z, w)

Processing without control inputs.

Return type:

Tensor

dymad.models.components.dyn_graph_direct(net, s, z, w)

Processing by GNN.

Return type:

Tensor

dymad.models.components.dyn_graph_skip(net, s, z, w)

Processing by GNN with skip connection.

Return type:

Tensor

dymad.models.components.dyn_skip(net, s, z, w)

Processing with skip connection.

Return type:

Tensor

dymad.models.components.enc_graph_auto(net, w)

Using GNN in EncAuto.

Return type:

Tensor

dymad.models.components.enc_graph_ctrl(net, w)

Using GNN in EncCtrl.

Return type:

Tensor

dymad.models.components.enc_graph_iden(net, w)

Identity encoder function for graph data.

Return type:

Tensor

dymad.models.components.enc_iden(net, w)

Identity encoder function.

Return type:

Tensor

dymad.models.components.enc_node_auto(net, w)

Using EncAuto for each node of graph.

Return type:

Tensor

dymad.models.components.enc_node_ctrl(net, w)

Using EncCtrl for each node of graph.

Return type:

Tensor

dymad.models.components.enc_node_raw_ctrl(net, w)

Using EncCtrl for each node of graph, letting encoder handle the concatenation.

Return type:

Tensor

dymad.models.components.enc_raw_ctrl(net, w)

Let encoder handle states and controls.

Return type:

Tensor

dymad.models.components.enc_smpl_auto(net, w)

Only encodes states.

Return type:

Tensor

dymad.models.components.enc_smpl_ctrl(net, w)

Encodes states and controls.

Return type:

Tensor

dymad.models.components.linear_eval_graph(mdl, w)

Compute linear evaluation, dz, and states, z, for the model.

Return type:

tuple[Tensor, Tensor]

dymad.models.components.linear_eval_smpl(mdl, w)

Compute linear evaluation, dz, and states, z, for the model.

Return type:

tuple[Tensor, Tensor]

dymad.models.components.linear_features_graph(mdl, w)

Compute linear features, f, and outputs, dz, for the model.

Return type:

tuple[Tensor, Tensor]

dymad.models.components.linear_features_smpl(mdl, w)

Compute linear features, f, and outputs, dz, for the model.

Return type:

tuple[Tensor, Tensor]

dymad.models.components.zu_blin_no_const(z, w)

Compute bilinear features without constant term.

Return type:

Tensor

dymad.models.components.zu_blin_no_const_graph(z, w)

Compute bilinear features without constant term for graph data.

Return type:

Tensor

dymad.models.components.zu_blin_with_const(z, w)

Compute bilinear features with constant term.

Return type:

Tensor

dymad.models.components.zu_blin_with_const_graph(z, w)

Compute bilinear features with constant term for graph data.

Return type:

Tensor

dymad.models.components.zu_cat_none(z, w)

No concatenation, just return z.

Return type:

Tensor

dymad.models.components.zu_cat_smpl(z, w)

Simple concatenation of z and u.

Return type:

Tensor

dymad.models.components.zu_cat_smpl_graph(z, w)

Simple concatenation of z and u on graph.

Return type:

Tensor